422
548
ie.revision = parent_entry.revision
423
549
return self._get_delta(ie, basis_inv, path), False, None
424
550
ie.reference_revision = content_summary[3]
426
self._add_text_to_weave(ie.file_id, lines, heads, None)
551
if ie.reference_revision is None:
552
raise AssertionError("invalid content_summary for nested tree: %r"
553
% (content_summary,))
554
self._add_text_to_weave(ie.file_id, '', heads, None)
428
556
raise NotImplementedError('unknown kind')
429
557
ie.revision = self._new_revision_id
558
self._any_changes = True
430
559
return self._get_delta(ie, basis_inv, path), True, fingerprint
432
def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
433
# Note: as we read the content directly from the tree, we know its not
434
# been turned into unicode or badly split - but a broken tree
435
# implementation could give us bad output from readlines() so this is
436
# not a guarantee of safety. What would be better is always checking
437
# the content during test suite execution. RBC 20070912
438
parent_keys = tuple((file_id, parent) for parent in parents)
439
return self.repository.texts.add_lines(
440
(file_id, self._new_revision_id), parent_keys, new_lines,
441
nostore_sha=nostore_sha, random_id=self.random_revid,
442
check_content=False)[0:2]
561
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
562
_entry_factory=entry_factory):
563
"""Record a new tree via iter_changes.
565
:param tree: The tree to obtain text contents from for changed objects.
566
:param basis_revision_id: The revision id of the tree the iter_changes
567
has been generated against. Currently assumed to be the same
568
as self.parents[0] - if it is not, errors may occur.
569
:param iter_changes: An iter_changes iterator with the changes to apply
570
to basis_revision_id. The iterator must not include any items with
571
a current kind of None - missing items must be either filtered out
572
or errored-on beefore record_iter_changes sees the item.
573
:param _entry_factory: Private method to bind entry_factory locally for
575
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
578
# Create an inventory delta based on deltas between all the parents and
579
# deltas between all the parent inventories. We use inventory delta's
580
# between the inventory objects because iter_changes masks
581
# last-changed-field only changes.
583
# file_id -> change map, change is fileid, paths, changed, versioneds,
584
# parents, names, kinds, executables
586
# {file_id -> revision_id -> inventory entry, for entries in parent
587
# trees that are not parents[0]
591
revtrees = list(self.repository.revision_trees(self.parents))
592
except errors.NoSuchRevision:
593
# one or more ghosts, slow path.
595
for revision_id in self.parents:
597
revtrees.append(self.repository.revision_tree(revision_id))
598
except errors.NoSuchRevision:
600
basis_revision_id = _mod_revision.NULL_REVISION
602
revtrees.append(self.repository.revision_tree(
603
_mod_revision.NULL_REVISION))
604
# The basis inventory from a repository
606
basis_inv = revtrees[0].inventory
608
basis_inv = self.repository.revision_tree(
609
_mod_revision.NULL_REVISION).inventory
610
if len(self.parents) > 0:
611
if basis_revision_id != self.parents[0] and not ghost_basis:
613
"arbitrary basis parents not yet supported with merges")
614
for revtree in revtrees[1:]:
615
for change in revtree.inventory._make_delta(basis_inv):
616
if change[1] is None:
617
# Not present in this parent.
619
if change[2] not in merged_ids:
620
if change[0] is not None:
621
basis_entry = basis_inv[change[2]]
622
merged_ids[change[2]] = [
624
basis_entry.revision,
627
parent_entries[change[2]] = {
629
basis_entry.revision:basis_entry,
631
change[3].revision:change[3],
634
merged_ids[change[2]] = [change[3].revision]
635
parent_entries[change[2]] = {change[3].revision:change[3]}
637
merged_ids[change[2]].append(change[3].revision)
638
parent_entries[change[2]][change[3].revision] = change[3]
641
# Setup the changes from the tree:
642
# changes maps file_id -> (change, [parent revision_ids])
644
for change in iter_changes:
645
# This probably looks up in basis_inv way to much.
646
if change[1][0] is not None:
647
head_candidate = [basis_inv[change[0]].revision]
650
changes[change[0]] = change, merged_ids.get(change[0],
652
unchanged_merged = set(merged_ids) - set(changes)
653
# Extend the changes dict with synthetic changes to record merges of
655
for file_id in unchanged_merged:
656
# Record a merged version of these items that did not change vs the
657
# basis. This can be either identical parallel changes, or a revert
658
# of a specific file after a merge. The recorded content will be
659
# that of the current tree (which is the same as the basis), but
660
# the per-file graph will reflect a merge.
661
# NB:XXX: We are reconstructing path information we had, this
662
# should be preserved instead.
663
# inv delta change: (file_id, (path_in_source, path_in_target),
664
# changed_content, versioned, parent, name, kind,
667
basis_entry = basis_inv[file_id]
668
except errors.NoSuchId:
669
# a change from basis->some_parents but file_id isn't in basis
670
# so was new in the merge, which means it must have changed
671
# from basis -> current, and as it hasn't the add was reverted
672
# by the user. So we discard this change.
676
(basis_inv.id2path(file_id), tree.id2path(file_id)),
678
(basis_entry.parent_id, basis_entry.parent_id),
679
(basis_entry.name, basis_entry.name),
680
(basis_entry.kind, basis_entry.kind),
681
(basis_entry.executable, basis_entry.executable))
682
changes[file_id] = (change, merged_ids[file_id])
683
# changes contains tuples with the change and a set of inventory
684
# candidates for the file.
686
# old_path, new_path, file_id, new_inventory_entry
687
seen_root = False # Is the root in the basis delta?
688
inv_delta = self._basis_delta
689
modified_rev = self._new_revision_id
690
for change, head_candidates in changes.values():
691
if change[3][1]: # versioned in target.
692
# Several things may be happening here:
693
# We may have a fork in the per-file graph
694
# - record a change with the content from tree
695
# We may have a change against < all trees
696
# - carry over the tree that hasn't changed
697
# We may have a change against all trees
698
# - record the change with the content from tree
701
entry = _entry_factory[kind](file_id, change[5][1],
703
head_set = self._heads(change[0], set(head_candidates))
706
for head_candidate in head_candidates:
707
if head_candidate in head_set:
708
heads.append(head_candidate)
709
head_set.remove(head_candidate)
712
# Could be a carry-over situation:
713
parent_entry_revs = parent_entries.get(file_id, None)
714
if parent_entry_revs:
715
parent_entry = parent_entry_revs.get(heads[0], None)
718
if parent_entry is None:
719
# The parent iter_changes was called against is the one
720
# that is the per-file head, so any change is relevant
721
# iter_changes is valid.
722
carry_over_possible = False
724
# could be a carry over situation
725
# A change against the basis may just indicate a merge,
726
# we need to check the content against the source of the
727
# merge to determine if it was changed after the merge
729
if (parent_entry.kind != entry.kind or
730
parent_entry.parent_id != entry.parent_id or
731
parent_entry.name != entry.name):
732
# Metadata common to all entries has changed
733
# against per-file parent
734
carry_over_possible = False
736
carry_over_possible = True
737
# per-type checks for changes against the parent_entry
740
# Cannot be a carry-over situation
741
carry_over_possible = False
742
# Populate the entry in the delta
744
# XXX: There is still a small race here: If someone reverts the content of a file
745
# after iter_changes examines and decides it has changed,
746
# we will unconditionally record a new version even if some
747
# other process reverts it while commit is running (with
748
# the revert happening after iter_changes did it's
751
entry.executable = True
753
entry.executable = False
754
if (carry_over_possible and
755
parent_entry.executable == entry.executable):
756
# Check the file length, content hash after reading
758
nostore_sha = parent_entry.text_sha1
761
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
763
text = file_obj.read()
767
entry.text_sha1, entry.text_size = self._add_text_to_weave(
768
file_id, text, heads, nostore_sha)
769
yield file_id, change[1][1], (entry.text_sha1, stat_value)
770
except errors.ExistingContent:
771
# No content change against a carry_over parent
772
# Perhaps this should also yield a fs hash update?
774
entry.text_size = parent_entry.text_size
775
entry.text_sha1 = parent_entry.text_sha1
776
elif kind == 'symlink':
778
entry.symlink_target = tree.get_symlink_target(file_id)
779
if (carry_over_possible and
780
parent_entry.symlink_target == entry.symlink_target):
783
self._add_text_to_weave(change[0], '', heads, None)
784
elif kind == 'directory':
785
if carry_over_possible:
788
# Nothing to set on the entry.
789
# XXX: split into the Root and nonRoot versions.
790
if change[1][1] != '' or self.repository.supports_rich_root():
791
self._add_text_to_weave(change[0], '', heads, None)
792
elif kind == 'tree-reference':
793
if not self.repository._format.supports_tree_reference:
794
# This isn't quite sane as an error, but we shouldn't
795
# ever see this code path in practice: tree's don't
796
# permit references when the repo doesn't support tree
798
raise errors.UnsupportedOperation(tree.add_reference,
800
reference_revision = tree.get_reference_revision(change[0])
801
entry.reference_revision = reference_revision
802
if (carry_over_possible and
803
parent_entry.reference_revision == reference_revision):
806
self._add_text_to_weave(change[0], '', heads, None)
808
raise AssertionError('unknown kind %r' % kind)
810
entry.revision = modified_rev
812
entry.revision = parent_entry.revision
815
new_path = change[1][1]
816
inv_delta.append((change[1][0], new_path, change[0], entry))
819
self.new_inventory = None
821
# This should perhaps be guarded by a check that the basis we
822
# commit against is the basis for the commit and if not do a delta
824
self._any_changes = True
826
# housekeeping root entry changes do not affect no-change commits.
827
self._require_root_change(tree)
828
self.basis_delta_revision = basis_revision_id
830
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
831
parent_keys = tuple([(file_id, parent) for parent in parents])
832
return self.repository.texts._add_text(
833
(file_id, self._new_revision_id), parent_keys, new_text,
834
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
445
837
class RootCommitBuilder(CommitBuilder):
446
838
"""This commitbuilder actually records the root id"""
448
840
# the root entry gets versioned properly by this builder.
449
841
_versioned_root = True
675
1168
# The old API returned a list, should this actually be a set?
676
1169
return parent_map.keys()
1171
def _check_inventories(self, checker):
1172
"""Check the inventories found from the revision scan.
1174
This is responsible for verifying the sha1 of inventories and
1175
creating a pending_keys set that covers data referenced by inventories.
1177
bar = ui.ui_factory.nested_progress_bar()
1179
self._do_check_inventories(checker, bar)
1183
def _do_check_inventories(self, checker, bar):
1184
"""Helper for _check_inventories."""
1186
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1187
kinds = ['chk_bytes', 'texts']
1188
count = len(checker.pending_keys)
1189
bar.update("inventories", 0, 2)
1190
current_keys = checker.pending_keys
1191
checker.pending_keys = {}
1192
# Accumulate current checks.
1193
for key in current_keys:
1194
if key[0] != 'inventories' and key[0] not in kinds:
1195
checker._report_items.append('unknown key type %r' % (key,))
1196
keys[key[0]].add(key[1:])
1197
if keys['inventories']:
1198
# NB: output order *should* be roughly sorted - topo or
1199
# inverse topo depending on repository - either way decent
1200
# to just delta against. However, pre-CHK formats didn't
1201
# try to optimise inventory layout on disk. As such the
1202
# pre-CHK code path does not use inventory deltas.
1204
for record in self.inventories.check(keys=keys['inventories']):
1205
if record.storage_kind == 'absent':
1206
checker._report_items.append(
1207
'Missing inventory {%s}' % (record.key,))
1209
last_object = self._check_record('inventories', record,
1210
checker, last_object,
1211
current_keys[('inventories',) + record.key])
1212
del keys['inventories']
1215
bar.update("texts", 1)
1216
while (checker.pending_keys or keys['chk_bytes']
1218
# Something to check.
1219
current_keys = checker.pending_keys
1220
checker.pending_keys = {}
1221
# Accumulate current checks.
1222
for key in current_keys:
1223
if key[0] not in kinds:
1224
checker._report_items.append('unknown key type %r' % (key,))
1225
keys[key[0]].add(key[1:])
1226
# Check the outermost kind only - inventories || chk_bytes || texts
1230
for record in getattr(self, kind).check(keys=keys[kind]):
1231
if record.storage_kind == 'absent':
1232
checker._report_items.append(
1233
'Missing %s {%s}' % (kind, record.key,))
1235
last_object = self._check_record(kind, record,
1236
checker, last_object, current_keys[(kind,) + record.key])
1240
def _check_record(self, kind, record, checker, last_object, item_data):
1241
"""Check a single text from this repository."""
1242
if kind == 'inventories':
1243
rev_id = record.key[0]
1244
inv = self._deserialise_inventory(rev_id,
1245
record.get_bytes_as('fulltext'))
1246
if last_object is not None:
1247
delta = inv._make_delta(last_object)
1248
for old_path, path, file_id, ie in delta:
1251
ie.check(checker, rev_id, inv)
1253
for path, ie in inv.iter_entries():
1254
ie.check(checker, rev_id, inv)
1255
if self._format.fast_deltas:
1257
elif kind == 'chk_bytes':
1258
# No code written to check chk_bytes for this repo format.
1259
checker._report_items.append(
1260
'unsupported key type chk_bytes for %s' % (record.key,))
1261
elif kind == 'texts':
1262
self._check_text(record, checker, item_data)
1264
checker._report_items.append(
1265
'unknown key type %s for %s' % (kind, record.key))
1267
def _check_text(self, record, checker, item_data):
1268
"""Check a single text."""
1269
# Check it is extractable.
1270
# TODO: check length.
1271
if record.storage_kind == 'chunked':
1272
chunks = record.get_bytes_as(record.storage_kind)
1273
sha1 = osutils.sha_strings(chunks)
1274
length = sum(map(len, chunks))
1276
content = record.get_bytes_as('fulltext')
1277
sha1 = osutils.sha_string(content)
1278
length = len(content)
1279
if item_data and sha1 != item_data[1]:
1280
checker._report_items.append(
1281
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1282
(record.key, sha1, item_data[1], item_data[2]))
679
1285
def create(a_bzrdir):
680
1286
"""Construct the current default format repository in a_bzrdir."""
952
1547
"""Commit the contents accrued within the current write group.
954
1549
:seealso: start_write_group.
1551
:return: it may return an opaque hint that can be passed to 'pack'.
956
1553
if self._write_group is not self.get_transaction():
957
1554
# has an unlock or relock occured ?
958
1555
raise errors.BzrError('mismatched lock context %r and '
959
1556
'write group %r.' %
960
1557
(self.get_transaction(), self._write_group))
961
self._commit_write_group()
1558
result = self._commit_write_group()
962
1559
self._write_group = None
964
1562
def _commit_write_group(self):
965
1563
"""Template method for per-repository write group cleanup.
967
This is called before the write group is considered to be
1565
This is called before the write group is considered to be
968
1566
finished and should ensure that all data handed to the repository
969
for writing during the write group is safely committed (to the
1567
for writing during the write group is safely committed (to the
970
1568
extent possible considering file system caching etc).
973
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1571
def suspend_write_group(self):
1572
raise errors.UnsuspendableWriteGroup(self)
1574
def get_missing_parent_inventories(self, check_for_missing_texts=True):
1575
"""Return the keys of missing inventory parents for revisions added in
1578
A revision is not complete if the inventory delta for that revision
1579
cannot be calculated. Therefore if the parent inventories of a
1580
revision are not present, the revision is incomplete, and e.g. cannot
1581
be streamed by a smart server. This method finds missing inventory
1582
parents for revisions added in this write group.
1584
if not self._format.supports_external_lookups:
1585
# This is only an issue for stacked repositories
1587
if not self.is_in_write_group():
1588
raise AssertionError('not in a write group')
1590
# XXX: We assume that every added revision already has its
1591
# corresponding inventory, so we only check for parent inventories that
1592
# might be missing, rather than all inventories.
1593
parents = set(self.revisions._index.get_missing_parents())
1594
parents.discard(_mod_revision.NULL_REVISION)
1595
unstacked_inventories = self.inventories._index
1596
present_inventories = unstacked_inventories.get_parent_map(
1597
key[-1:] for key in parents)
1598
parents.difference_update(present_inventories)
1599
if len(parents) == 0:
1600
# No missing parent inventories.
1602
if not check_for_missing_texts:
1603
return set(('inventories', rev_id) for (rev_id,) in parents)
1604
# Ok, now we have a list of missing inventories. But these only matter
1605
# if the inventories that reference them are missing some texts they
1606
# appear to introduce.
1607
# XXX: Texts referenced by all added inventories need to be present,
1608
# but at the moment we're only checking for texts referenced by
1609
# inventories at the graph's edge.
1610
key_deps = self.revisions._index._key_dependencies
1611
key_deps.satisfy_refs_for_keys(present_inventories)
1612
referrers = frozenset(r[0] for r in key_deps.get_referrers())
1613
file_ids = self.fileids_altered_by_revision_ids(referrers)
1614
missing_texts = set()
1615
for file_id, version_ids in file_ids.iteritems():
1616
missing_texts.update(
1617
(file_id, version_id) for version_id in version_ids)
1618
present_texts = self.texts.get_parent_map(missing_texts)
1619
missing_texts.difference_update(present_texts)
1620
if not missing_texts:
1621
# No texts are missing, so all revisions and their deltas are
1624
# Alternatively the text versions could be returned as the missing
1625
# keys, but this is likely to be less data.
1626
missing_keys = set(('inventories', rev_id) for (rev_id,) in parents)
1629
def refresh_data(self):
1630
"""Re-read any data needed to to synchronise with disk.
1632
This method is intended to be called after another repository instance
1633
(such as one used by a smart server) has inserted data into the
1634
repository. It may not be called during a write group, but may be
1635
called at any other time.
1637
if self.is_in_write_group():
1638
raise errors.InternalBzrError(
1639
"May not refresh_data while in a write group.")
1640
self._refresh_data()
1642
def resume_write_group(self, tokens):
1643
if not self.is_write_locked():
1644
raise errors.NotWriteLocked(self)
1645
if self._write_group:
1646
raise errors.BzrError('already in a write group')
1647
self._resume_write_group(tokens)
1648
# so we can detect unlock/relock - the write group is now entered.
1649
self._write_group = self.get_transaction()
1651
def _resume_write_group(self, tokens):
1652
raise errors.UnsuspendableWriteGroup(self)
1654
def fetch(self, source, revision_id=None, pb=None, find_ghosts=False,
974
1656
"""Fetch the content required to construct revision_id from source.
976
If revision_id is None all content is copied.
1658
If revision_id is None and fetch_spec is None, then all content is
1661
fetch() may not be used when the repository is in a write group -
1662
either finish the current write group before using fetch, or use
1663
fetch before starting the write group.
977
1665
:param find_ghosts: Find and copy revisions in the source that are
978
1666
ghosts in the target (and not reachable directly by walking out to
979
1667
the first-present revision in target from revision_id).
1668
:param revision_id: If specified, all the content needed for this
1669
revision ID will be copied to the target. Fetch will determine for
1670
itself which content needs to be copied.
1671
:param fetch_spec: If specified, a SearchResult or
1672
PendingAncestryResult that describes which revisions to copy. This
1673
allows copying multiple heads at once. Mutually exclusive with
1676
if fetch_spec is not None and revision_id is not None:
1677
raise AssertionError(
1678
"fetch_spec and revision_id are mutually exclusive.")
1679
if self.is_in_write_group():
1680
raise errors.InternalBzrError(
1681
"May not fetch while in a write group.")
981
1682
# fast path same-url fetch operations
982
if self.has_same_location(source):
1683
# TODO: lift out to somewhere common with RemoteRepository
1684
# <https://bugs.edge.launchpad.net/bzr/+bug/401646>
1685
if (self.has_same_location(source)
1686
and fetch_spec is None
1687
and self._has_same_fallbacks(source)):
983
1688
# check that last_revision is in 'from' and then return a
985
1690
if (revision_id is not None and
1133
1853
@needs_read_lock
1134
1854
def get_revisions(self, revision_ids):
1135
"""Get many revisions at once."""
1855
"""Get many revisions at once.
1857
Repositories that need to check data on every revision read should
1858
subclass this method.
1136
1860
return self._get_revisions(revision_ids)
1138
1862
@needs_read_lock
1139
1863
def _get_revisions(self, revision_ids):
1140
1864
"""Core work logic to get many revisions without sanity checks."""
1141
for rev_id in revision_ids:
1142
if not rev_id or not isinstance(rev_id, basestring):
1143
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1866
for revid, rev in self._iter_revisions(revision_ids):
1868
raise errors.NoSuchRevision(self, revid)
1870
return [revs[revid] for revid in revision_ids]
1872
def _iter_revisions(self, revision_ids):
1873
"""Iterate over revision objects.
1875
:param revision_ids: An iterable of revisions to examine. None may be
1876
passed to request all revisions known to the repository. Note that
1877
not all repositories can find unreferenced revisions; for those
1878
repositories only referenced ones will be returned.
1879
:return: An iterator of (revid, revision) tuples. Absent revisions (
1880
those asked for but not available) are returned as (revid, None).
1882
if revision_ids is None:
1883
revision_ids = self.all_revision_ids()
1885
for rev_id in revision_ids:
1886
if not rev_id or not isinstance(rev_id, basestring):
1887
raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
1144
1888
keys = [(key,) for key in revision_ids]
1145
1889
stream = self.revisions.get_record_stream(keys, 'unordered', True)
1147
1890
for record in stream:
1891
revid = record.key[0]
1148
1892
if record.storage_kind == 'absent':
1149
raise errors.NoSuchRevision(self, record.key[0])
1150
text = record.get_bytes_as('fulltext')
1151
rev = self._serializer.read_revision_from_string(text)
1152
revs[record.key[0]] = rev
1153
return [revs[revid] for revid in revision_ids]
1156
def get_revision_xml(self, revision_id):
1157
# TODO: jam 20070210 This shouldn't be necessary since get_revision
1158
# would have already do it.
1159
# TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1160
rev = self.get_revision(revision_id)
1161
rev_tmp = StringIO()
1162
# the current serializer..
1163
self._serializer.write_revision(rev, rev_tmp)
1165
return rev_tmp.getvalue()
1167
def get_deltas_for_revisions(self, revisions):
1895
text = record.get_bytes_as('fulltext')
1896
rev = self._serializer.read_revision_from_string(text)
1899
def get_deltas_for_revisions(self, revisions, specific_fileids=None):
1168
1900
"""Produce a generator of revision deltas.
1170
1902
Note that the input is a sequence of REVISIONS, not revision_ids.
1171
1903
Trees will be held in memory until the generator exits.
1172
1904
Each delta is relative to the revision's lefthand predecessor.
1906
:param specific_fileids: if not None, the result is filtered
1907
so that only those file-ids, their parents and their
1908
children are included.
1910
# Get the revision-ids of interest
1174
1911
required_trees = set()
1175
1912
for revision in revisions:
1176
1913
required_trees.add(revision.revision_id)
1177
1914
required_trees.update(revision.parent_ids[:1])
1178
trees = dict((t.get_revision_id(), t) for
1179
t in self.revision_trees(required_trees))
1916
# Get the matching filtered trees. Note that it's more
1917
# efficient to pass filtered trees to changes_from() rather
1918
# than doing the filtering afterwards. changes_from() could
1919
# arguably do the filtering itself but it's path-based, not
1920
# file-id based, so filtering before or afterwards is
1922
if specific_fileids is None:
1923
trees = dict((t.get_revision_id(), t) for
1924
t in self.revision_trees(required_trees))
1926
trees = dict((t.get_revision_id(), t) for
1927
t in self._filtered_revision_trees(required_trees,
1930
# Calculate the deltas
1180
1931
for revision in revisions:
1181
1932
if not revision.parent_ids:
1182
1933
old_tree = self.revision_tree(_mod_revision.NULL_REVISION)
1538
2340
"""Get Inventory object by revision id."""
1539
2341
return self.iter_inventories([revision_id]).next()
1541
def iter_inventories(self, revision_ids):
2343
def iter_inventories(self, revision_ids, ordering=None):
1542
2344
"""Get many inventories by revision_ids.
1544
2346
This will buffer some or all of the texts used in constructing the
1545
2347
inventories in memory, but will only parse a single inventory at a
2350
:param revision_ids: The expected revision ids of the inventories.
2351
:param ordering: optional ordering, e.g. 'topological'. If not
2352
specified, the order of revision_ids will be preserved (by
2353
buffering if necessary).
1548
2354
:return: An iterator of inventories.
1550
2356
if ((None in revision_ids)
1551
2357
or (_mod_revision.NULL_REVISION in revision_ids)):
1552
2358
raise ValueError('cannot get null revision inventory')
1553
return self._iter_inventories(revision_ids)
2359
return self._iter_inventories(revision_ids, ordering)
1555
def _iter_inventories(self, revision_ids):
2361
def _iter_inventories(self, revision_ids, ordering):
1556
2362
"""single-document based inventory iteration."""
1557
for text, revision_id in self._iter_inventory_xmls(revision_ids):
1558
yield self.deserialise_inventory(revision_id, text)
2363
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2364
for text, revision_id in inv_xmls:
2365
yield self._deserialise_inventory(revision_id, text)
1560
def _iter_inventory_xmls(self, revision_ids):
2367
def _iter_inventory_xmls(self, revision_ids, ordering):
2368
if ordering is None:
2369
order_as_requested = True
2370
ordering = 'unordered'
2372
order_as_requested = False
1561
2373
keys = [(revision_id,) for revision_id in revision_ids]
1562
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2376
if order_as_requested:
2377
key_iter = iter(keys)
2378
next_key = key_iter.next()
2379
stream = self.inventories.get_record_stream(keys, ordering, True)
1564
2381
for record in stream:
1565
2382
if record.storage_kind != 'absent':
1566
texts[record.key] = record.get_bytes_as('fulltext')
2383
chunks = record.get_bytes_as('chunked')
2384
if order_as_requested:
2385
text_chunks[record.key] = chunks
2387
yield ''.join(chunks), record.key[-1]
1568
2389
raise errors.NoSuchRevision(self, record.key)
1570
yield texts[key], key[-1]
2390
if order_as_requested:
2391
# Yield as many results as we can while preserving order.
2392
while next_key in text_chunks:
2393
chunks = text_chunks.pop(next_key)
2394
yield ''.join(chunks), next_key[-1]
2396
next_key = key_iter.next()
2397
except StopIteration:
2398
# We still want to fully consume the get_record_stream,
2399
# just in case it is not actually finished at this point
1572
def deserialise_inventory(self, revision_id, xml):
1573
"""Transform the xml into an inventory object.
2403
def _deserialise_inventory(self, revision_id, xml):
2404
"""Transform the xml into an inventory object.
1575
2406
:param revision_id: The expected revision id of the inventory.
1576
2407
:param xml: A serialised inventory.
1578
result = self._serializer.read_inventory_from_string(xml, revision_id)
2409
result = self._serializer.read_inventory_from_string(xml, revision_id,
2410
entry_cache=self._inventory_entry_cache,
2411
return_from_cache=self._safe_to_return_from_cache)
1579
2412
if result.revision_id != revision_id:
1580
2413
raise AssertionError('revision id mismatch %s != %s' % (
1581
2414
result.revision_id, revision_id))
1584
def serialise_inventory(self, inv):
1585
return self._serializer.write_inventory_to_string(inv)
1587
def _serialise_inventory_to_lines(self, inv):
1588
return self._serializer.write_inventory_to_lines(inv)
1590
2417
def get_serializer_format(self):
1591
2418
return self._serializer.format_num
1593
2420
@needs_read_lock
1594
def get_inventory_xml(self, revision_id):
1595
"""Get inventory XML as a file object."""
1596
texts = self._iter_inventory_xmls([revision_id])
2421
def _get_inventory_xml(self, revision_id):
2422
"""Get serialized inventory as a string."""
2423
texts = self._iter_inventory_xmls([revision_id], 'unordered')
1598
2425
text, revision_id = texts.next()
1599
2426
except StopIteration:
1600
2427
raise errors.HistoryMissing(self, 'inventory', revision_id)
1604
def get_inventory_sha1(self, revision_id):
1605
"""Return the sha1 hash of the inventory entry
2430
def get_rev_id_for_revno(self, revno, known_pair):
2431
"""Return the revision id of a revno, given a later (revno, revid)
2432
pair in the same history.
2434
:return: if found (True, revid). If the available history ran out
2435
before reaching the revno, then this returns
2436
(False, (closest_revno, closest_revid)).
1607
return self.get_revision(revision_id).inventory_sha1
2438
known_revno, known_revid = known_pair
2439
partial_history = [known_revid]
2440
distance_from_known = known_revno - revno
2441
if distance_from_known < 0:
2443
'requested revno (%d) is later than given known revno (%d)'
2444
% (revno, known_revno))
2447
self, partial_history, stop_index=distance_from_known)
2448
except errors.RevisionNotPresent, err:
2449
if err.revision_id == known_revid:
2450
# The start revision (known_revid) wasn't found.
2452
# This is a stacked repository with no fallbacks, or a there's a
2453
# left-hand ghost. Either way, even though the revision named in
2454
# the error isn't in this repo, we know it's the next step in this
2455
# left-hand history.
2456
partial_history.append(err.revision_id)
2457
if len(partial_history) <= distance_from_known:
2458
# Didn't find enough history to get a revid for the revno.
2459
earliest_revno = known_revno - len(partial_history) + 1
2460
return (False, (earliest_revno, partial_history[-1]))
2461
if len(partial_history) - 1 > distance_from_known:
2462
raise AssertionError('_iter_for_revno returned too much history')
2463
return (True, partial_history[-1])
1609
2465
def iter_reverse_revision_history(self, revision_id):
1610
2466
"""Iterate backwards through revision ids in the lefthand history
2737
3690
return self.source.revision_ids_to_search_result(result_set)
2740
class InterPackRepo(InterSameDataRepository):
2741
"""Optimised code paths between Pack based repositories."""
2744
def _get_repo_format_to_test(self):
2745
from bzrlib.repofmt import pack_repo
2746
return pack_repo.RepositoryFormatKnitPack1()
2749
def is_compatible(source, target):
2750
"""Be compatible with known Pack formats.
2752
We don't test for the stores being of specific types because that
2753
could lead to confusing results, and there is no need to be
2756
from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2758
are_packs = (isinstance(source._format, RepositoryFormatPack) and
2759
isinstance(target._format, RepositoryFormatPack))
2760
except AttributeError:
2762
return are_packs and InterRepository._same_model(source, target)
2765
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2766
"""See InterRepository.fetch()."""
2767
if (len(self.source._fallback_repositories) > 0 or
2768
len(self.target._fallback_repositories) > 0):
2769
# The pack layer is not aware of fallback repositories, so when
2770
# fetching from a stacked repository or into a stacked repository
2771
# we use the generic fetch logic which uses the VersionedFiles
2772
# attributes on repository.
2773
from bzrlib.fetch import RepoFetcher
2774
fetcher = RepoFetcher(self.target, self.source, revision_id,
2776
return fetcher.count_copied, fetcher.failed_revisions
2777
from bzrlib.repofmt.pack_repo import Packer
2778
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2779
self.source, self.source._format, self.target, self.target._format)
2780
self.count_copied = 0
2781
if revision_id is None:
2783
# everything to do - use pack logic
2784
# to fetch from all packs to one without
2785
# inventory parsing etc, IFF nothing to be copied is in the target.
2787
source_revision_ids = frozenset(self.source.all_revision_ids())
2788
revision_ids = source_revision_ids - \
2789
frozenset(self.target.get_parent_map(source_revision_ids))
2790
revision_keys = [(revid,) for revid in revision_ids]
2791
index = self.target._pack_collection.revision_index.combined_index
2792
present_revision_ids = set(item[1][0] for item in
2793
index.iter_entries(revision_keys))
2794
revision_ids = set(revision_ids) - present_revision_ids
2795
# implementing the TODO will involve:
2796
# - detecting when all of a pack is selected
2797
# - avoiding as much as possible pre-selection, so the
2798
# more-core routines such as create_pack_from_packs can filter in
2799
# a just-in-time fashion. (though having a HEADS list on a
2800
# repository might make this a lot easier, because we could
2801
# sensibly detect 'new revisions' without doing a full index scan.
2802
elif _mod_revision.is_null(revision_id):
2807
revision_ids = self.search_missing_revision_ids(revision_id,
2808
find_ghosts=find_ghosts).get_keys()
2809
except errors.NoSuchRevision:
2810
raise errors.InstallFailed([revision_id])
2811
if len(revision_ids) == 0:
2813
packs = self.source._pack_collection.all_packs()
2814
pack = Packer(self.target._pack_collection, packs, '.fetch',
2815
revision_ids).pack()
2816
if pack is not None:
2817
self.target._pack_collection._save_pack_names()
2818
# Trigger an autopack. This may duplicate effort as we've just done
2819
# a pack creation, but for now it is simpler to think about as
2820
# 'upload data, then repack if needed'.
2821
self.target._pack_collection.autopack()
2822
return (pack.get_revision_count(), [])
2827
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
2828
"""See InterRepository.missing_revision_ids().
2830
:param find_ghosts: Find ghosts throughout the ancestry of
2833
if not find_ghosts and revision_id is not None:
2834
return self._walk_to_common_revisions([revision_id])
2835
elif revision_id is not None:
2836
# Find ghosts: search for revisions pointing from one repository to
2837
# the other, and vice versa, anywhere in the history of revision_id.
2838
graph = self.target.get_graph(other_repository=self.source)
2839
searcher = graph._make_breadth_first_searcher([revision_id])
2843
next_revs, ghosts = searcher.next_with_ghosts()
2844
except StopIteration:
2846
if revision_id in ghosts:
2847
raise errors.NoSuchRevision(self.source, revision_id)
2848
found_ids.update(next_revs)
2849
found_ids.update(ghosts)
2850
found_ids = frozenset(found_ids)
2851
# Double query here: should be able to avoid this by changing the
2852
# graph api further.
2853
result_set = found_ids - frozenset(
2854
self.target.get_parent_map(found_ids))
2856
source_ids = self.source.all_revision_ids()
2857
# source_ids is the worst possible case we may need to pull.
2858
# now we want to filter source_ids against what we actually
2859
# have in target, but don't try to check for existence where we know
2860
# we do not have a revision as that would be pointless.
2861
target_ids = set(self.target.all_revision_ids())
2862
result_set = set(source_ids).difference(target_ids)
2863
return self.source.revision_ids_to_search_result(result_set)
2866
class InterModel1and2(InterRepository):
2869
def _get_repo_format_to_test(self):
2873
def is_compatible(source, target):
2874
if not source.supports_rich_root() and target.supports_rich_root():
2880
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2881
"""See InterRepository.fetch()."""
2882
from bzrlib.fetch import Model1toKnit2Fetcher
2883
f = Model1toKnit2Fetcher(to_repository=self.target,
2884
from_repository=self.source,
2885
last_revision=revision_id,
2886
pb=pb, find_ghosts=find_ghosts)
2887
return f.count_copied, f.failed_revisions
2890
def copy_content(self, revision_id=None):
2891
"""Make a complete copy of the content in self into destination.
2893
This is a destructive operation! Do not use it on existing
2896
:param revision_id: Only copy the content needed to construct
2897
revision_id and its parents.
2900
self.target.set_make_working_trees(self.source.make_working_trees())
2901
except NotImplementedError:
2903
# but don't bother fetching if we have the needed data now.
2904
if (revision_id not in (None, _mod_revision.NULL_REVISION) and
2905
self.target.has_revision(revision_id)):
2907
self.target.fetch(self.source, revision_id=revision_id)
2910
class InterKnit1and2(InterKnitRepo):
2913
def _get_repo_format_to_test(self):
2917
def is_compatible(source, target):
2918
"""Be compatible with Knit1 source and Knit3 target"""
2920
from bzrlib.repofmt.knitrepo import (
2921
RepositoryFormatKnit1,
2922
RepositoryFormatKnit3,
2924
from bzrlib.repofmt.pack_repo import (
2925
RepositoryFormatKnitPack1,
2926
RepositoryFormatKnitPack3,
2927
RepositoryFormatKnitPack4,
2928
RepositoryFormatKnitPack5,
2929
RepositoryFormatKnitPack5RichRoot,
2930
RepositoryFormatPackDevelopment2,
2931
RepositoryFormatPackDevelopment2Subtree,
2934
RepositoryFormatKnit1, # no rr, no subtree
2935
RepositoryFormatKnitPack1, # no rr, no subtree
2936
RepositoryFormatPackDevelopment2, # no rr, no subtree
2937
RepositoryFormatKnitPack5, # no rr, no subtree
2940
RepositoryFormatKnit3, # rr, subtree
2941
RepositoryFormatKnitPack3, # rr, subtree
2942
RepositoryFormatKnitPack4, # rr, no subtree
2943
RepositoryFormatKnitPack5RichRoot,# rr, no subtree
2944
RepositoryFormatPackDevelopment2Subtree, # rr, subtree
2946
for format in norichroot:
2947
if format.rich_root_data:
2948
raise AssertionError('Format %s is a rich-root format'
2949
' but is included in the non-rich-root list'
2951
for format in richroot:
2952
if not format.rich_root_data:
2953
raise AssertionError('Format %s is not a rich-root format'
2954
' but is included in the rich-root list'
2956
# TODO: One alternative is to just check format.rich_root_data,
2957
# instead of keeping membership lists. However, the formats
2958
# *also* have to use the same 'Knit' style of storage
2959
# (line-deltas, fulltexts, etc.)
2960
return (isinstance(source._format, norichroot) and
2961
isinstance(target._format, richroot))
2962
except AttributeError:
2966
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2967
"""See InterRepository.fetch()."""
2968
from bzrlib.fetch import Knit1to2Fetcher
2969
mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2970
self.source, self.source._format, self.target,
2971
self.target._format)
2972
f = Knit1to2Fetcher(to_repository=self.target,
2973
from_repository=self.source,
2974
last_revision=revision_id,
2975
pb=pb, find_ghosts=find_ghosts)
2976
return f.count_copied, f.failed_revisions
2979
class InterDifferingSerializer(InterKnitRepo):
3693
class InterDifferingSerializer(InterRepository):
2982
3696
def _get_repo_format_to_test(self):
2986
3700
def is_compatible(source, target):
2987
3701
"""Be compatible with Knit2 source and Knit3 target"""
2988
if source.supports_rich_root() != target.supports_rich_root():
2990
# Ideally, we'd support fetching if the source had no tree references
2991
# even if it supported them...
2992
if (getattr(source, '_format.supports_tree_reference', False) and
2993
not getattr(target, '_format.supports_tree_reference', False)):
3702
# This is redundant with format.check_conversion_target(), however that
3703
# raises an exception, and we just want to say "False" as in we won't
3704
# support converting between these formats.
3705
if 'IDS_never' in debug.debug_flags:
3707
if source.supports_rich_root() and not target.supports_rich_root():
3709
if (source._format.supports_tree_reference
3710
and not target._format.supports_tree_reference):
3712
if target._fallback_repositories and target._format.supports_chks:
3713
# IDS doesn't know how to copy CHKs for the parent inventories it
3714
# adds to stacked repos.
3716
if 'IDS_always' in debug.debug_flags:
3718
# Only use this code path for local source and target. IDS does far
3719
# too much IO (both bandwidth and roundtrips) over a network.
3720
if not source.bzrdir.transport.base.startswith('file:///'):
3722
if not target.bzrdir.transport.base.startswith('file:///'):
3726
def _get_trees(self, revision_ids, cache):
3728
for rev_id in revision_ids:
3730
possible_trees.append((rev_id, cache[rev_id]))
3732
# Not cached, but inventory might be present anyway.
3734
tree = self.source.revision_tree(rev_id)
3735
except errors.NoSuchRevision:
3736
# Nope, parent is ghost.
3739
cache[rev_id] = tree
3740
possible_trees.append((rev_id, tree))
3741
return possible_trees
3743
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3744
"""Get the best delta and base for this revision.
3746
:return: (basis_id, delta)
3749
# Generate deltas against each tree, to find the shortest.
3750
texts_possibly_new_in_tree = set()
3751
for basis_id, basis_tree in possible_trees:
3752
delta = tree.inventory._make_delta(basis_tree.inventory)
3753
for old_path, new_path, file_id, new_entry in delta:
3754
if new_path is None:
3755
# This file_id isn't present in the new rev, so we don't
3759
# Rich roots are handled elsewhere...
3761
kind = new_entry.kind
3762
if kind != 'directory' and kind != 'file':
3763
# No text record associated with this inventory entry.
3765
# This is a directory or file that has changed somehow.
3766
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3767
deltas.append((len(delta), basis_id, delta))
3769
return deltas[0][1:]
3771
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3772
"""Find all parent revisions that are absent, but for which the
3773
inventory is present, and copy those inventories.
3775
This is necessary to preserve correctness when the source is stacked
3776
without fallbacks configured. (Note that in cases like upgrade the
3777
source may be not have _fallback_repositories even though it is
3781
for parents in parent_map.values():
3782
parent_revs.update(parents)
3783
present_parents = self.source.get_parent_map(parent_revs)
3784
absent_parents = set(parent_revs).difference(present_parents)
3785
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3786
(rev_id,) for rev_id in absent_parents)
3787
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3788
for parent_tree in self.source.revision_trees(parent_inv_ids):
3789
current_revision_id = parent_tree.get_revision_id()
3790
parents_parents_keys = parent_invs_keys_for_stacking[
3791
(current_revision_id,)]
3792
parents_parents = [key[-1] for key in parents_parents_keys]
3793
basis_id = _mod_revision.NULL_REVISION
3794
basis_tree = self.source.revision_tree(basis_id)
3795
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3796
self.target.add_inventory_by_delta(
3797
basis_id, delta, current_revision_id, parents_parents)
3798
cache[current_revision_id] = parent_tree
3800
def _fetch_batch(self, revision_ids, basis_id, cache, a_graph=None):
3801
"""Fetch across a few revisions.
3803
:param revision_ids: The revisions to copy
3804
:param basis_id: The revision_id of a tree that must be in cache, used
3805
as a basis for delta when no other base is available
3806
:param cache: A cache of RevisionTrees that we can use.
3807
:param a_graph: A Graph object to determine the heads() of the
3808
rich-root data stream.
3809
:return: The revision_id of the last converted tree. The RevisionTree
3810
for it will be in cache
3812
# Walk though all revisions; get inventory deltas, copy referenced
3813
# texts that delta references, insert the delta, revision and
3815
root_keys_to_create = set()
3818
pending_revisions = []
3819
parent_map = self.source.get_parent_map(revision_ids)
3820
self._fetch_parent_invs_for_stacking(parent_map, cache)
3821
self.source._safe_to_return_from_cache = True
3822
for tree in self.source.revision_trees(revision_ids):
3823
# Find a inventory delta for this revision.
3824
# Find text entries that need to be copied, too.
3825
current_revision_id = tree.get_revision_id()
3826
parent_ids = parent_map.get(current_revision_id, ())
3827
parent_trees = self._get_trees(parent_ids, cache)
3828
possible_trees = list(parent_trees)
3829
if len(possible_trees) == 0:
3830
# There either aren't any parents, or the parents are ghosts,
3831
# so just use the last converted tree.
3832
possible_trees.append((basis_id, cache[basis_id]))
3833
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3835
revision = self.source.get_revision(current_revision_id)
3836
pending_deltas.append((basis_id, delta,
3837
current_revision_id, revision.parent_ids))
3838
if self._converting_to_rich_root:
3839
self._revision_id_to_root_id[current_revision_id] = \
3841
# Determine which texts are in present in this revision but not in
3842
# any of the available parents.
3843
texts_possibly_new_in_tree = set()
3844
for old_path, new_path, file_id, entry in delta:
3845
if new_path is None:
3846
# This file_id isn't present in the new rev
3850
if not self.target.supports_rich_root():
3851
# The target doesn't support rich root, so we don't
3854
if self._converting_to_rich_root:
3855
# This can't be copied normally, we have to insert
3857
root_keys_to_create.add((file_id, entry.revision))
3860
texts_possibly_new_in_tree.add((file_id, entry.revision))
3861
for basis_id, basis_tree in possible_trees:
3862
basis_inv = basis_tree.inventory
3863
for file_key in list(texts_possibly_new_in_tree):
3864
file_id, file_revision = file_key
3866
entry = basis_inv[file_id]
3867
except errors.NoSuchId:
3869
if entry.revision == file_revision:
3870
texts_possibly_new_in_tree.remove(file_key)
3871
text_keys.update(texts_possibly_new_in_tree)
3872
pending_revisions.append(revision)
3873
cache[current_revision_id] = tree
3874
basis_id = current_revision_id
3875
self.source._safe_to_return_from_cache = False
3877
from_texts = self.source.texts
3878
to_texts = self.target.texts
3879
if root_keys_to_create:
3880
root_stream = _mod_fetch._new_root_data_stream(
3881
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3882
self.source, graph=a_graph)
3883
to_texts.insert_record_stream(root_stream)
3884
to_texts.insert_record_stream(from_texts.get_record_stream(
3885
text_keys, self.target._format._fetch_order,
3886
not self.target._format._fetch_uses_deltas))
3887
# insert inventory deltas
3888
for delta in pending_deltas:
3889
self.target.add_inventory_by_delta(*delta)
3890
if self.target._fallback_repositories:
3891
# Make sure this stacked repository has all the parent inventories
3892
# for the new revisions that we are about to insert. We do this
3893
# before adding the revisions so that no revision is added until
3894
# all the inventories it may depend on are added.
3895
# Note that this is overzealous, as we may have fetched these in an
3898
revision_ids = set()
3899
for revision in pending_revisions:
3900
revision_ids.add(revision.revision_id)
3901
parent_ids.update(revision.parent_ids)
3902
parent_ids.difference_update(revision_ids)
3903
parent_ids.discard(_mod_revision.NULL_REVISION)
3904
parent_map = self.source.get_parent_map(parent_ids)
3905
# we iterate over parent_map and not parent_ids because we don't
3906
# want to try copying any revision which is a ghost
3907
for parent_tree in self.source.revision_trees(parent_map):
3908
current_revision_id = parent_tree.get_revision_id()
3909
parents_parents = parent_map[current_revision_id]
3910
possible_trees = self._get_trees(parents_parents, cache)
3911
if len(possible_trees) == 0:
3912
# There either aren't any parents, or the parents are
3913
# ghosts, so just use the last converted tree.
3914
possible_trees.append((basis_id, cache[basis_id]))
3915
basis_id, delta = self._get_delta_for_revision(parent_tree,
3916
parents_parents, possible_trees)
3917
self.target.add_inventory_by_delta(
3918
basis_id, delta, current_revision_id, parents_parents)
3919
# insert signatures and revisions
3920
for revision in pending_revisions:
3922
signature = self.source.get_signature_text(
3923
revision.revision_id)
3924
self.target.add_signature_text(revision.revision_id,
3926
except errors.NoSuchRevision:
3928
self.target.add_revision(revision.revision_id, revision)
3931
def _fetch_all_revisions(self, revision_ids, pb):
3932
"""Fetch everything for the list of revisions.
3934
:param revision_ids: The list of revisions to fetch. Must be in
3936
:param pb: A ProgressTask
3939
basis_id, basis_tree = self._get_basis(revision_ids[0])
3941
cache = lru_cache.LRUCache(100)
3942
cache[basis_id] = basis_tree
3943
del basis_tree # We don't want to hang on to it here
3945
if self._converting_to_rich_root and len(revision_ids) > 100:
3946
a_graph = _mod_fetch._get_rich_root_heads_graph(self.source,
3951
for offset in range(0, len(revision_ids), batch_size):
3952
self.target.start_write_group()
3954
pb.update('Transferring revisions', offset,
3956
batch = revision_ids[offset:offset+batch_size]
3957
basis_id = self._fetch_batch(batch, basis_id, cache,
3960
self.source._safe_to_return_from_cache = False
3961
self.target.abort_write_group()
3964
hint = self.target.commit_write_group()
3967
if hints and self.target._format.pack_compresses:
3968
self.target.pack(hint=hints)
3969
pb.update('Transferring revisions', len(revision_ids),
2997
3972
@needs_write_lock
2998
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3973
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
2999
3975
"""See InterRepository.fetch()."""
3976
if fetch_spec is not None:
3977
raise AssertionError("Not implemented yet...")
3978
ui.ui_factory.warn_experimental_format_fetch(self)
3979
if (not self.source.supports_rich_root()
3980
and self.target.supports_rich_root()):
3981
self._converting_to_rich_root = True
3982
self._revision_id_to_root_id = {}
3984
self._converting_to_rich_root = False
3985
# See <https://launchpad.net/bugs/456077> asking for a warning here
3986
if self.source._format.network_name() != self.target._format.network_name():
3987
ui.ui_factory.show_user_warning('cross_format_fetch',
3988
from_format=self.source._format,
3989
to_format=self.target._format)
3000
3990
revision_ids = self.target.search_missing_revision_ids(self.source,
3001
3991
revision_id, find_ghosts=find_ghosts).get_keys()
3992
if not revision_ids:
3002
3994
revision_ids = tsort.topo_sort(
3003
3995
self.source.get_graph().get_parent_map(revision_ids))
3004
def revisions_iterator():
3005
for current_revision_id in revision_ids:
3006
revision = self.source.get_revision(current_revision_id)
3007
tree = self.source.revision_tree(current_revision_id)
3009
signature = self.source.get_signature_text(
3010
current_revision_id)
3011
except errors.NoSuchRevision:
3013
yield revision, tree, signature
3996
if not revision_ids:
3998
# Walk though all revisions; get inventory deltas, copy referenced
3999
# texts that delta references, insert the delta, revision and
3015
4002
my_pb = ui.ui_factory.nested_progress_bar()
4005
symbol_versioning.warn(
4006
symbol_versioning.deprecated_in((1, 14, 0))
4007
% "pb parameter to fetch()")
3020
install_revisions(self.target, revisions_iterator(),
3021
len(revision_ids), pb)
4010
self._fetch_all_revisions(revision_ids, pb)
3023
4012
if my_pb is not None:
3024
4013
my_pb.finished()
3025
4014
return len(revision_ids), 0
3028
class InterOtherToRemote(InterRepository):
3030
def __init__(self, source, target):
3031
InterRepository.__init__(self, source, target)
3032
self._real_inter = None
3035
def is_compatible(source, target):
3036
if isinstance(target, remote.RemoteRepository):
3040
def _ensure_real_inter(self):
3041
if self._real_inter is None:
3042
self.target._ensure_real()
3043
real_target = self.target._real_repository
3044
self._real_inter = InterRepository.get(self.source, real_target)
3046
def copy_content(self, revision_id=None):
3047
self._ensure_real_inter()
3048
self._real_inter.copy_content(revision_id=revision_id)
3050
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3051
self._ensure_real_inter()
3052
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3053
find_ghosts=find_ghosts)
3056
def _get_repo_format_to_test(self):
3060
class InterRemoteToOther(InterRepository):
3062
def __init__(self, source, target):
3063
InterRepository.__init__(self, source, target)
3064
self._real_inter = None
3067
def is_compatible(source, target):
3068
if not isinstance(source, remote.RemoteRepository):
3070
# Is source's model compatible with target's model?
3071
source._ensure_real()
3072
real_source = source._real_repository
3073
if isinstance(real_source, remote.RemoteRepository):
3074
raise NotImplementedError(
3075
"We don't support remote repos backed by remote repos yet.")
3076
return InterRepository._same_model(real_source, target)
3078
def _ensure_real_inter(self):
3079
if self._real_inter is None:
3080
self.source._ensure_real()
3081
real_source = self.source._real_repository
3082
self._real_inter = InterRepository.get(real_source, self.target)
3084
def fetch(self, revision_id=None, pb=None, find_ghosts=False):
3085
self._ensure_real_inter()
3086
return self._real_inter.fetch(revision_id=revision_id, pb=pb,
3087
find_ghosts=find_ghosts)
3089
def copy_content(self, revision_id=None):
3090
self._ensure_real_inter()
3091
self._real_inter.copy_content(revision_id=revision_id)
3094
def _get_repo_format_to_test(self):
4016
def _get_basis(self, first_revision_id):
4017
"""Get a revision and tree which exists in the target.
4019
This assumes that first_revision_id is selected for transmission
4020
because all other ancestors are already present. If we can't find an
4021
ancestor we fall back to NULL_REVISION since we know that is safe.
4023
:return: (basis_id, basis_tree)
4025
first_rev = self.source.get_revision(first_revision_id)
4027
basis_id = first_rev.parent_ids[0]
4028
# only valid as a basis if the target has it
4029
self.target.get_revision(basis_id)
4030
# Try to get a basis tree - if its a ghost it will hit the
4031
# NoSuchRevision case.
4032
basis_tree = self.source.revision_tree(basis_id)
4033
except (IndexError, errors.NoSuchRevision):
4034
basis_id = _mod_revision.NULL_REVISION
4035
basis_tree = self.source.revision_tree(basis_id)
4036
return basis_id, basis_tree
3099
4039
InterRepository.register_optimiser(InterDifferingSerializer)
3100
4040
InterRepository.register_optimiser(InterSameDataRepository)
3101
4041
InterRepository.register_optimiser(InterWeaveRepo)
3102
4042
InterRepository.register_optimiser(InterKnitRepo)
3103
InterRepository.register_optimiser(InterModel1and2)
3104
InterRepository.register_optimiser(InterKnit1and2)
3105
InterRepository.register_optimiser(InterPackRepo)
3106
InterRepository.register_optimiser(InterOtherToRemote)
3107
InterRepository.register_optimiser(InterRemoteToOther)
3110
4045
class CopyConverter(object):
3111
4046
"""A repository conversion tool which just performs a copy of the content.
3113
4048
This is slow but quite reliable.
3261
4201
revision_graph[key] = tuple(parent for parent in parents if parent
3262
4202
in revision_graph)
3263
4203
return revision_graph
4206
class StreamSink(object):
4207
"""An object that can insert a stream into a repository.
4209
This interface handles the complexity of reserialising inventories and
4210
revisions from different formats, and allows unidirectional insertion into
4211
stacked repositories without looking for the missing basis parents
4215
def __init__(self, target_repo):
4216
self.target_repo = target_repo
4218
def insert_stream(self, stream, src_format, resume_tokens):
4219
"""Insert a stream's content into the target repository.
4221
:param src_format: a bzr repository format.
4223
:return: a list of resume tokens and an iterable of keys additional
4224
items required before the insertion can be completed.
4226
self.target_repo.lock_write()
4229
self.target_repo.resume_write_group(resume_tokens)
4232
self.target_repo.start_write_group()
4235
# locked_insert_stream performs a commit|suspend.
4236
return self._locked_insert_stream(stream, src_format, is_resume)
4238
self.target_repo.abort_write_group(suppress_errors=True)
4241
self.target_repo.unlock()
4243
def _locked_insert_stream(self, stream, src_format, is_resume):
4244
to_serializer = self.target_repo._format._serializer
4245
src_serializer = src_format._serializer
4247
if to_serializer == src_serializer:
4248
# If serializers match and the target is a pack repository, set the
4249
# write cache size on the new pack. This avoids poor performance
4250
# on transports where append is unbuffered (such as
4251
# RemoteTransport). This is safe to do because nothing should read
4252
# back from the target repository while a stream with matching
4253
# serialization is being inserted.
4254
# The exception is that a delta record from the source that should
4255
# be a fulltext may need to be expanded by the target (see
4256
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4257
# explicitly flush any buffered writes first in that rare case.
4259
new_pack = self.target_repo._pack_collection._new_pack
4260
except AttributeError:
4261
# Not a pack repository
4264
new_pack.set_write_cache_size(1024*1024)
4265
for substream_type, substream in stream:
4266
if 'stream' in debug.debug_flags:
4267
mutter('inserting substream: %s', substream_type)
4268
if substream_type == 'texts':
4269
self.target_repo.texts.insert_record_stream(substream)
4270
elif substream_type == 'inventories':
4271
if src_serializer == to_serializer:
4272
self.target_repo.inventories.insert_record_stream(
4275
self._extract_and_insert_inventories(
4276
substream, src_serializer)
4277
elif substream_type == 'inventory-deltas':
4278
self._extract_and_insert_inventory_deltas(
4279
substream, src_serializer)
4280
elif substream_type == 'chk_bytes':
4281
# XXX: This doesn't support conversions, as it assumes the
4282
# conversion was done in the fetch code.
4283
self.target_repo.chk_bytes.insert_record_stream(substream)
4284
elif substream_type == 'revisions':
4285
# This may fallback to extract-and-insert more often than
4286
# required if the serializers are different only in terms of
4288
if src_serializer == to_serializer:
4289
self.target_repo.revisions.insert_record_stream(
4292
self._extract_and_insert_revisions(substream,
4294
elif substream_type == 'signatures':
4295
self.target_repo.signatures.insert_record_stream(substream)
4297
raise AssertionError('kaboom! %s' % (substream_type,))
4298
# Done inserting data, and the missing_keys calculations will try to
4299
# read back from the inserted data, so flush the writes to the new pack
4300
# (if this is pack format).
4301
if new_pack is not None:
4302
new_pack._write_data('', flush=True)
4303
# Find all the new revisions (including ones from resume_tokens)
4304
missing_keys = self.target_repo.get_missing_parent_inventories(
4305
check_for_missing_texts=is_resume)
4307
for prefix, versioned_file in (
4308
('texts', self.target_repo.texts),
4309
('inventories', self.target_repo.inventories),
4310
('revisions', self.target_repo.revisions),
4311
('signatures', self.target_repo.signatures),
4312
('chk_bytes', self.target_repo.chk_bytes),
4314
if versioned_file is None:
4316
# TODO: key is often going to be a StaticTuple object
4317
# I don't believe we can define a method by which
4318
# (prefix,) + StaticTuple will work, though we could
4319
# define a StaticTuple.sq_concat that would allow you to
4320
# pass in either a tuple or a StaticTuple as the second
4321
# object, so instead we could have:
4322
# StaticTuple(prefix) + key here...
4323
missing_keys.update((prefix,) + key for key in
4324
versioned_file.get_missing_compression_parent_keys())
4325
except NotImplementedError:
4326
# cannot even attempt suspending, and missing would have failed
4327
# during stream insertion.
4328
missing_keys = set()
4331
# suspend the write group and tell the caller what we is
4332
# missing. We know we can suspend or else we would not have
4333
# entered this code path. (All repositories that can handle
4334
# missing keys can handle suspending a write group).
4335
write_group_tokens = self.target_repo.suspend_write_group()
4336
return write_group_tokens, missing_keys
4337
hint = self.target_repo.commit_write_group()
4338
if (to_serializer != src_serializer and
4339
self.target_repo._format.pack_compresses):
4340
self.target_repo.pack(hint=hint)
4343
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4344
target_rich_root = self.target_repo._format.rich_root_data
4345
target_tree_refs = self.target_repo._format.supports_tree_reference
4346
for record in substream:
4347
# Insert the delta directly
4348
inventory_delta_bytes = record.get_bytes_as('fulltext')
4349
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4351
parse_result = deserialiser.parse_text_bytes(
4352
inventory_delta_bytes)
4353
except inventory_delta.IncompatibleInventoryDelta, err:
4354
trace.mutter("Incompatible delta: %s", err.msg)
4355
raise errors.IncompatibleRevision(self.target_repo._format)
4356
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4357
revision_id = new_id
4358
parents = [key[0] for key in record.parents]
4359
self.target_repo.add_inventory_by_delta(
4360
basis_id, inv_delta, revision_id, parents)
4362
def _extract_and_insert_inventories(self, substream, serializer,
4364
"""Generate a new inventory versionedfile in target, converting data.
4366
The inventory is retrieved from the source, (deserializing it), and
4367
stored in the target (reserializing it in a different format).
4369
target_rich_root = self.target_repo._format.rich_root_data
4370
target_tree_refs = self.target_repo._format.supports_tree_reference
4371
for record in substream:
4372
# It's not a delta, so it must be a fulltext in the source
4373
# serializer's format.
4374
bytes = record.get_bytes_as('fulltext')
4375
revision_id = record.key[0]
4376
inv = serializer.read_inventory_from_string(bytes, revision_id)
4377
parents = [key[0] for key in record.parents]
4378
self.target_repo.add_inventory(revision_id, inv, parents)
4379
# No need to keep holding this full inv in memory when the rest of
4380
# the substream is likely to be all deltas.
4383
def _extract_and_insert_revisions(self, substream, serializer):
4384
for record in substream:
4385
bytes = record.get_bytes_as('fulltext')
4386
revision_id = record.key[0]
4387
rev = serializer.read_revision_from_string(bytes)
4388
if rev.revision_id != revision_id:
4389
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4390
self.target_repo.add_revision(revision_id, rev)
4393
if self.target_repo._format._fetch_reconcile:
4394
self.target_repo.reconcile()
4397
class StreamSource(object):
4398
"""A source of a stream for fetching between repositories."""
4400
def __init__(self, from_repository, to_format):
4401
"""Create a StreamSource streaming from from_repository."""
4402
self.from_repository = from_repository
4403
self.to_format = to_format
4405
def delta_on_metadata(self):
4406
"""Return True if delta's are permitted on metadata streams.
4408
That is on revisions and signatures.
4410
src_serializer = self.from_repository._format._serializer
4411
target_serializer = self.to_format._serializer
4412
return (self.to_format._fetch_uses_deltas and
4413
src_serializer == target_serializer)
4415
def _fetch_revision_texts(self, revs):
4416
# fetch signatures first and then the revision texts
4417
# may need to be a InterRevisionStore call here.
4418
from_sf = self.from_repository.signatures
4419
# A missing signature is just skipped.
4420
keys = [(rev_id,) for rev_id in revs]
4421
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4423
self.to_format._fetch_order,
4424
not self.to_format._fetch_uses_deltas))
4425
# If a revision has a delta, this is actually expanded inside the
4426
# insert_record_stream code now, which is an alternate fix for
4428
from_rf = self.from_repository.revisions
4429
revisions = from_rf.get_record_stream(
4431
self.to_format._fetch_order,
4432
not self.delta_on_metadata())
4433
return [('signatures', signatures), ('revisions', revisions)]
4435
def _generate_root_texts(self, revs):
4436
"""This will be called by get_stream between fetching weave texts and
4437
fetching the inventory weave.
4439
if self._rich_root_upgrade():
4440
return _mod_fetch.Inter1and2Helper(
4441
self.from_repository).generate_root_texts(revs)
4445
def get_stream(self, search):
4447
revs = search.get_keys()
4448
graph = self.from_repository.get_graph()
4449
revs = tsort.topo_sort(graph.get_parent_map(revs))
4450
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4452
for knit_kind, file_id, revisions in data_to_fetch:
4453
if knit_kind != phase:
4455
# Make a new progress bar for this phase
4456
if knit_kind == "file":
4457
# Accumulate file texts
4458
text_keys.extend([(file_id, revision) for revision in
4460
elif knit_kind == "inventory":
4461
# Now copy the file texts.
4462
from_texts = self.from_repository.texts
4463
yield ('texts', from_texts.get_record_stream(
4464
text_keys, self.to_format._fetch_order,
4465
not self.to_format._fetch_uses_deltas))
4466
# Cause an error if a text occurs after we have done the
4469
# Before we process the inventory we generate the root
4470
# texts (if necessary) so that the inventories references
4472
for _ in self._generate_root_texts(revs):
4474
# we fetch only the referenced inventories because we do not
4475
# know for unselected inventories whether all their required
4476
# texts are present in the other repository - it could be
4478
for info in self._get_inventory_stream(revs):
4480
elif knit_kind == "signatures":
4481
# Nothing to do here; this will be taken care of when
4482
# _fetch_revision_texts happens.
4484
elif knit_kind == "revisions":
4485
for record in self._fetch_revision_texts(revs):
4488
raise AssertionError("Unknown knit kind %r" % knit_kind)
4490
def get_stream_for_missing_keys(self, missing_keys):
4491
# missing keys can only occur when we are byte copying and not
4492
# translating (because translation means we don't send
4493
# unreconstructable deltas ever).
4495
keys['texts'] = set()
4496
keys['revisions'] = set()
4497
keys['inventories'] = set()
4498
keys['chk_bytes'] = set()
4499
keys['signatures'] = set()
4500
for key in missing_keys:
4501
keys[key[0]].add(key[1:])
4502
if len(keys['revisions']):
4503
# If we allowed copying revisions at this point, we could end up
4504
# copying a revision without copying its required texts: a
4505
# violation of the requirements for repository integrity.
4506
raise AssertionError(
4507
'cannot copy revisions to fill in missing deltas %s' % (
4508
keys['revisions'],))
4509
for substream_kind, keys in keys.iteritems():
4510
vf = getattr(self.from_repository, substream_kind)
4511
if vf is None and keys:
4512
raise AssertionError(
4513
"cannot fill in keys for a versioned file we don't"
4514
" have: %s needs %s" % (substream_kind, keys))
4516
# No need to stream something we don't have
4518
if substream_kind == 'inventories':
4519
# Some missing keys are genuinely ghosts, filter those out.
4520
present = self.from_repository.inventories.get_parent_map(keys)
4521
revs = [key[0] for key in present]
4522
# Get the inventory stream more-or-less as we do for the
4523
# original stream; there's no reason to assume that records
4524
# direct from the source will be suitable for the sink. (Think
4525
# e.g. 2a -> 1.9-rich-root).
4526
for info in self._get_inventory_stream(revs, missing=True):
4530
# Ask for full texts always so that we don't need more round trips
4531
# after this stream.
4532
# Some of the missing keys are genuinely ghosts, so filter absent
4533
# records. The Sink is responsible for doing another check to
4534
# ensure that ghosts don't introduce missing data for future
4536
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4537
self.to_format._fetch_order, True))
4538
yield substream_kind, stream
4540
def inventory_fetch_order(self):
4541
if self._rich_root_upgrade():
4542
return 'topological'
4544
return self.to_format._fetch_order
4546
def _rich_root_upgrade(self):
4547
return (not self.from_repository._format.rich_root_data and
4548
self.to_format.rich_root_data)
4550
def _get_inventory_stream(self, revision_ids, missing=False):
4551
from_format = self.from_repository._format
4552
if (from_format.supports_chks and self.to_format.supports_chks and
4553
from_format.network_name() == self.to_format.network_name()):
4554
raise AssertionError(
4555
"this case should be handled by GroupCHKStreamSource")
4556
elif 'forceinvdeltas' in debug.debug_flags:
4557
return self._get_convertable_inventory_stream(revision_ids,
4558
delta_versus_null=missing)
4559
elif from_format.network_name() == self.to_format.network_name():
4561
return self._get_simple_inventory_stream(revision_ids,
4563
elif (not from_format.supports_chks and not self.to_format.supports_chks
4564
and from_format._serializer == self.to_format._serializer):
4565
# Essentially the same format.
4566
return self._get_simple_inventory_stream(revision_ids,
4569
# Any time we switch serializations, we want to use an
4570
# inventory-delta based approach.
4571
return self._get_convertable_inventory_stream(revision_ids,
4572
delta_versus_null=missing)
4574
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4575
# NB: This currently reopens the inventory weave in source;
4576
# using a single stream interface instead would avoid this.
4577
from_weave = self.from_repository.inventories
4579
delta_closure = True
4581
delta_closure = not self.delta_on_metadata()
4582
yield ('inventories', from_weave.get_record_stream(
4583
[(rev_id,) for rev_id in revision_ids],
4584
self.inventory_fetch_order(), delta_closure))
4586
def _get_convertable_inventory_stream(self, revision_ids,
4587
delta_versus_null=False):
4588
# The two formats are sufficiently different that there is no fast
4589
# path, so we need to send just inventorydeltas, which any
4590
# sufficiently modern client can insert into any repository.
4591
# The StreamSink code expects to be able to
4592
# convert on the target, so we need to put bytes-on-the-wire that can
4593
# be converted. That means inventory deltas (if the remote is <1.19,
4594
# RemoteStreamSink will fallback to VFS to insert the deltas).
4595
yield ('inventory-deltas',
4596
self._stream_invs_as_deltas(revision_ids,
4597
delta_versus_null=delta_versus_null))
4599
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4600
"""Return a stream of inventory-deltas for the given rev ids.
4602
:param revision_ids: The list of inventories to transmit
4603
:param delta_versus_null: Don't try to find a minimal delta for this
4604
entry, instead compute the delta versus the NULL_REVISION. This
4605
effectively streams a complete inventory. Used for stuff like
4606
filling in missing parents, etc.
4608
from_repo = self.from_repository
4609
revision_keys = [(rev_id,) for rev_id in revision_ids]
4610
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4611
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4613
inventories = self.from_repository.iter_inventories(
4614
revision_ids, 'topological')
4615
format = from_repo._format
4616
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4617
inventory_cache = lru_cache.LRUCache(50)
4618
null_inventory = from_repo.revision_tree(
4619
_mod_revision.NULL_REVISION).inventory
4620
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4621
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4622
# repo back into a non-rich-root repo ought to be allowed)
4623
serializer = inventory_delta.InventoryDeltaSerializer(
4624
versioned_root=format.rich_root_data,
4625
tree_references=format.supports_tree_reference)
4626
for inv in inventories:
4627
key = (inv.revision_id,)
4628
parent_keys = parent_map.get(key, ())
4630
if not delta_versus_null and parent_keys:
4631
# The caller did not ask for complete inventories and we have
4632
# some parents that we can delta against. Make a delta against
4633
# each parent so that we can find the smallest.
4634
parent_ids = [parent_key[0] for parent_key in parent_keys]
4635
for parent_id in parent_ids:
4636
if parent_id not in invs_sent_so_far:
4637
# We don't know that the remote side has this basis, so
4640
if parent_id == _mod_revision.NULL_REVISION:
4641
parent_inv = null_inventory
4643
parent_inv = inventory_cache.get(parent_id, None)
4644
if parent_inv is None:
4645
parent_inv = from_repo.get_inventory(parent_id)
4646
candidate_delta = inv._make_delta(parent_inv)
4647
if (delta is None or
4648
len(delta) > len(candidate_delta)):
4649
delta = candidate_delta
4650
basis_id = parent_id
4652
# Either none of the parents ended up being suitable, or we
4653
# were asked to delta against NULL
4654
basis_id = _mod_revision.NULL_REVISION
4655
delta = inv._make_delta(null_inventory)
4656
invs_sent_so_far.add(inv.revision_id)
4657
inventory_cache[inv.revision_id] = inv
4658
delta_serialized = ''.join(
4659
serializer.delta_to_lines(basis_id, key[-1], delta))
4660
yield versionedfile.FulltextContentFactory(
4661
key, parent_keys, None, delta_serialized)
4664
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4665
stop_revision=None):
4666
"""Extend the partial history to include a given index
4668
If a stop_index is supplied, stop when that index has been reached.
4669
If a stop_revision is supplied, stop when that revision is
4670
encountered. Otherwise, stop when the beginning of history is
4673
:param stop_index: The index which should be present. When it is
4674
present, history extension will stop.
4675
:param stop_revision: The revision id which should be present. When
4676
it is encountered, history extension will stop.
4678
start_revision = partial_history_cache[-1]
4679
iterator = repo.iter_reverse_revision_history(start_revision)
4681
#skip the last revision in the list
4684
if (stop_index is not None and
4685
len(partial_history_cache) > stop_index):
4687
if partial_history_cache[-1] == stop_revision:
4689
revision_id = iterator.next()
4690
partial_history_cache.append(revision_id)
4691
except StopIteration:
added
removed
bzrlib/repository.py
