353
208
commit to be valid, deletes against the basis MUST be recorded via
354
209
builder.record_delete().
356
self._recording_deletes = True
358
basis_id = self.parents[0]
360
basis_id = _mod_revision.NULL_REVISION
361
self.basis_delta_revision = basis_id
363
def record_entry_contents(self, ie, parent_invs, path, tree,
365
"""Record the content of ie from tree into the commit if needed.
367
Side effect: sets ie.revision when unchanged
369
:param ie: An inventory entry present in the commit.
370
:param parent_invs: The inventories of the parent revisions of the
372
:param path: The path the entry is at in the tree.
373
:param tree: The tree which contains this entry and should be used to
375
:param content_summary: Summary data from the tree about the paths
376
content - stat, length, exec, sha/link target. This is only
377
accessed when the entry has a revision of None - that is when it is
378
a candidate to commit.
379
:return: A tuple (change_delta, version_recorded, fs_hash).
380
change_delta is an inventory_delta change for this entry against
381
the basis tree of the commit, or None if no change occured against
383
version_recorded is True if a new version of the entry has been
384
recorded. For instance, committing a merge where a file was only
385
changed on the other side will return (delta, False).
386
fs_hash is either None, or the hash details for the path (currently
387
a tuple of the contents sha1 and the statvalue returned by
388
tree.get_file_with_stat()).
390
if self.new_inventory.root is None:
391
if ie.parent_id is not None:
392
raise errors.RootMissing()
393
self._check_root(ie, parent_invs, tree)
394
if ie.revision is None:
395
kind = content_summary[0]
397
# ie is carried over from a prior commit
399
# XXX: repository specific check for nested tree support goes here - if
400
# the repo doesn't want nested trees we skip it ?
401
if (kind == 'tree-reference' and
402
not self.repository._format.supports_tree_reference):
403
# mismatch between commit builder logic and repository:
404
# this needs the entry creation pushed down into the builder.
405
raise NotImplementedError('Missing repository subtree support.')
406
self.new_inventory.add(ie)
408
# TODO: slow, take it out of the inner loop.
410
basis_inv = parent_invs[0]
412
basis_inv = Inventory(root_id=None)
414
# ie.revision is always None if the InventoryEntry is considered
415
# for committing. We may record the previous parents revision if the
416
# content is actually unchanged against a sole head.
417
if ie.revision is not None:
418
if not self._versioned_root and path == '':
419
# repositories that do not version the root set the root's
420
# revision to the new commit even when no change occurs (more
421
# specifically, they do not record a revision on the root; and
422
# the rev id is assigned to the root during deserialisation -
423
# this masks when a change may have occurred against the basis.
424
# To match this we always issue a delta, because the revision
425
# of the root will always be changing.
426
if ie.file_id in basis_inv:
427
delta = (basis_inv.id2path(ie.file_id), path,
431
delta = (None, path, ie.file_id, ie)
432
self._basis_delta.append(delta)
433
return delta, False, None
435
# we don't need to commit this, because the caller already
436
# determined that an existing revision of this file is
437
# appropriate. If it's not being considered for committing then
438
# it and all its parents to the root must be unaltered so
439
# no-change against the basis.
440
if ie.revision == self._new_revision_id:
441
raise AssertionError("Impossible situation, a skipped "
442
"inventory entry (%r) claims to be modified in this "
443
"commit (%r).", (ie, self._new_revision_id))
444
return None, False, None
445
# XXX: Friction: parent_candidates should return a list not a dict
446
# so that we don't have to walk the inventories again.
447
parent_candiate_entries = ie.parent_candidates(parent_invs)
448
head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
450
for inv in parent_invs:
451
if ie.file_id in inv:
452
old_rev = inv[ie.file_id].revision
453
if old_rev in head_set:
454
heads.append(inv[ie.file_id].revision)
455
head_set.remove(inv[ie.file_id].revision)
458
# now we check to see if we need to write a new record to the
460
# We write a new entry unless there is one head to the ancestors, and
461
# the kind-derived content is unchanged.
463
# Cheapest check first: no ancestors, or more the one head in the
464
# ancestors, we write a new node.
468
# There is a single head, look it up for comparison
469
parent_entry = parent_candiate_entries[heads[0]]
470
# if the non-content specific data has changed, we'll be writing a
472
if (parent_entry.parent_id != ie.parent_id or
473
parent_entry.name != ie.name):
475
# now we need to do content specific checks:
477
# if the kind changed the content obviously has
478
if kind != parent_entry.kind:
480
# Stat cache fingerprint feedback for the caller - None as we usually
481
# don't generate one.
484
if content_summary[2] is None:
485
raise ValueError("Files must not have executable = None")
487
# We can't trust a check of the file length because of content
489
if (# if the exec bit has changed we have to store:
490
parent_entry.executable != content_summary[2]):
492
elif parent_entry.text_sha1 == content_summary[3]:
493
# all meta and content is unchanged (using a hash cache
494
# hit to check the sha)
495
ie.revision = parent_entry.revision
496
ie.text_size = parent_entry.text_size
497
ie.text_sha1 = parent_entry.text_sha1
498
ie.executable = parent_entry.executable
499
return self._get_delta(ie, basis_inv, path), False, None
501
# Either there is only a hash change(no hash cache entry,
502
# or same size content change), or there is no change on
504
# Provide the parent's hash to the store layer, so that the
505
# content is unchanged we will not store a new node.
506
nostore_sha = parent_entry.text_sha1
508
# We want to record a new node regardless of the presence or
509
# absence of a content change in the file.
511
ie.executable = content_summary[2]
512
file_obj, stat_value = tree.get_file_with_stat(ie.file_id, path)
514
text = file_obj.read()
518
ie.text_sha1, ie.text_size = self._add_text_to_weave(
519
ie.file_id, text, heads, nostore_sha)
520
# Let the caller know we generated a stat fingerprint.
521
fingerprint = (ie.text_sha1, stat_value)
522
except errors.ExistingContent:
523
# Turns out that the file content was unchanged, and we were
524
# only going to store a new node if it was changed. Carry over
526
ie.revision = parent_entry.revision
527
ie.text_size = parent_entry.text_size
528
ie.text_sha1 = parent_entry.text_sha1
529
ie.executable = parent_entry.executable
530
return self._get_delta(ie, basis_inv, path), False, None
531
elif kind == 'directory':
533
# all data is meta here, nothing specific to directory, so
535
ie.revision = parent_entry.revision
536
return self._get_delta(ie, basis_inv, path), False, None
537
self._add_text_to_weave(ie.file_id, '', heads, None)
538
elif kind == 'symlink':
539
current_link_target = content_summary[3]
541
# symlink target is not generic metadata, check if it has
543
if current_link_target != parent_entry.symlink_target:
546
# unchanged, carry over.
547
ie.revision = parent_entry.revision
548
ie.symlink_target = parent_entry.symlink_target
549
return self._get_delta(ie, basis_inv, path), False, None
550
ie.symlink_target = current_link_target
551
self._add_text_to_weave(ie.file_id, '', heads, None)
552
elif kind == 'tree-reference':
554
if content_summary[3] != parent_entry.reference_revision:
557
# unchanged, carry over.
558
ie.reference_revision = parent_entry.reference_revision
559
ie.revision = parent_entry.revision
560
return self._get_delta(ie, basis_inv, path), False, None
561
ie.reference_revision = content_summary[3]
562
if ie.reference_revision is None:
563
raise AssertionError("invalid content_summary for nested tree: %r"
564
% (content_summary,))
565
self._add_text_to_weave(ie.file_id, '', heads, None)
567
raise NotImplementedError('unknown kind')
568
ie.revision = self._new_revision_id
569
self._any_changes = True
570
return self._get_delta(ie, basis_inv, path), True, fingerprint
572
def record_iter_changes(self, tree, basis_revision_id, iter_changes,
573
_entry_factory=entry_factory):
211
raise NotImplementedError(self.will_record_deletes)
213
def record_iter_changes(self, tree, basis_revision_id, iter_changes):
574
214
"""Record a new tree via iter_changes.
576
216
:param tree: The tree to obtain text contents from for changed objects.
581
221
to basis_revision_id. The iterator must not include any items with
582
222
a current kind of None - missing items must be either filtered out
583
223
or errored-on beefore record_iter_changes sees the item.
584
:param _entry_factory: Private method to bind entry_factory locally for
586
224
:return: A generator of (file_id, relpath, fs_hash) tuples for use with
587
225
tree._observed_sha1.
589
# Create an inventory delta based on deltas between all the parents and
590
# deltas between all the parent inventories. We use inventory delta's
591
# between the inventory objects because iter_changes masks
592
# last-changed-field only changes.
594
# file_id -> change map, change is fileid, paths, changed, versioneds,
595
# parents, names, kinds, executables
597
# {file_id -> revision_id -> inventory entry, for entries in parent
598
# trees that are not parents[0]
602
revtrees = list(self.repository.revision_trees(self.parents))
603
except errors.NoSuchRevision:
604
# one or more ghosts, slow path.
606
for revision_id in self.parents:
608
revtrees.append(self.repository.revision_tree(revision_id))
609
except errors.NoSuchRevision:
611
basis_revision_id = _mod_revision.NULL_REVISION
613
revtrees.append(self.repository.revision_tree(
614
_mod_revision.NULL_REVISION))
615
# The basis inventory from a repository
617
basis_inv = revtrees[0].inventory
619
basis_inv = self.repository.revision_tree(
620
_mod_revision.NULL_REVISION).inventory
621
if len(self.parents) > 0:
622
if basis_revision_id != self.parents[0] and not ghost_basis:
624
"arbitrary basis parents not yet supported with merges")
625
for revtree in revtrees[1:]:
626
for change in revtree.inventory._make_delta(basis_inv):
627
if change[1] is None:
628
# Not present in this parent.
630
if change[2] not in merged_ids:
631
if change[0] is not None:
632
basis_entry = basis_inv[change[2]]
633
merged_ids[change[2]] = [
635
basis_entry.revision,
638
parent_entries[change[2]] = {
640
basis_entry.revision:basis_entry,
642
change[3].revision:change[3],
645
merged_ids[change[2]] = [change[3].revision]
646
parent_entries[change[2]] = {change[3].revision:change[3]}
648
merged_ids[change[2]].append(change[3].revision)
649
parent_entries[change[2]][change[3].revision] = change[3]
652
# Setup the changes from the tree:
653
# changes maps file_id -> (change, [parent revision_ids])
655
for change in iter_changes:
656
# This probably looks up in basis_inv way to much.
657
if change[1][0] is not None:
658
head_candidate = [basis_inv[change[0]].revision]
661
changes[change[0]] = change, merged_ids.get(change[0],
663
unchanged_merged = set(merged_ids) - set(changes)
664
# Extend the changes dict with synthetic changes to record merges of
666
for file_id in unchanged_merged:
667
# Record a merged version of these items that did not change vs the
668
# basis. This can be either identical parallel changes, or a revert
669
# of a specific file after a merge. The recorded content will be
670
# that of the current tree (which is the same as the basis), but
671
# the per-file graph will reflect a merge.
672
# NB:XXX: We are reconstructing path information we had, this
673
# should be preserved instead.
674
# inv delta change: (file_id, (path_in_source, path_in_target),
675
# changed_content, versioned, parent, name, kind,
678
basis_entry = basis_inv[file_id]
679
except errors.NoSuchId:
680
# a change from basis->some_parents but file_id isn't in basis
681
# so was new in the merge, which means it must have changed
682
# from basis -> current, and as it hasn't the add was reverted
683
# by the user. So we discard this change.
687
(basis_inv.id2path(file_id), tree.id2path(file_id)),
689
(basis_entry.parent_id, basis_entry.parent_id),
690
(basis_entry.name, basis_entry.name),
691
(basis_entry.kind, basis_entry.kind),
692
(basis_entry.executable, basis_entry.executable))
693
changes[file_id] = (change, merged_ids[file_id])
694
# changes contains tuples with the change and a set of inventory
695
# candidates for the file.
697
# old_path, new_path, file_id, new_inventory_entry
698
seen_root = False # Is the root in the basis delta?
699
inv_delta = self._basis_delta
700
modified_rev = self._new_revision_id
701
for change, head_candidates in changes.values():
702
if change[3][1]: # versioned in target.
703
# Several things may be happening here:
704
# We may have a fork in the per-file graph
705
# - record a change with the content from tree
706
# We may have a change against < all trees
707
# - carry over the tree that hasn't changed
708
# We may have a change against all trees
709
# - record the change with the content from tree
712
entry = _entry_factory[kind](file_id, change[5][1],
714
head_set = self._heads(change[0], set(head_candidates))
717
for head_candidate in head_candidates:
718
if head_candidate in head_set:
719
heads.append(head_candidate)
720
head_set.remove(head_candidate)
723
# Could be a carry-over situation:
724
parent_entry_revs = parent_entries.get(file_id, None)
725
if parent_entry_revs:
726
parent_entry = parent_entry_revs.get(heads[0], None)
729
if parent_entry is None:
730
# The parent iter_changes was called against is the one
731
# that is the per-file head, so any change is relevant
732
# iter_changes is valid.
733
carry_over_possible = False
735
# could be a carry over situation
736
# A change against the basis may just indicate a merge,
737
# we need to check the content against the source of the
738
# merge to determine if it was changed after the merge
740
if (parent_entry.kind != entry.kind or
741
parent_entry.parent_id != entry.parent_id or
742
parent_entry.name != entry.name):
743
# Metadata common to all entries has changed
744
# against per-file parent
745
carry_over_possible = False
747
carry_over_possible = True
748
# per-type checks for changes against the parent_entry
751
# Cannot be a carry-over situation
752
carry_over_possible = False
753
# Populate the entry in the delta
755
# XXX: There is still a small race here: If someone reverts the content of a file
756
# after iter_changes examines and decides it has changed,
757
# we will unconditionally record a new version even if some
758
# other process reverts it while commit is running (with
759
# the revert happening after iter_changes did its
762
entry.executable = True
764
entry.executable = False
765
if (carry_over_possible and
766
parent_entry.executable == entry.executable):
767
# Check the file length, content hash after reading
769
nostore_sha = parent_entry.text_sha1
772
file_obj, stat_value = tree.get_file_with_stat(file_id, change[1][1])
774
text = file_obj.read()
778
entry.text_sha1, entry.text_size = self._add_text_to_weave(
779
file_id, text, heads, nostore_sha)
780
yield file_id, change[1][1], (entry.text_sha1, stat_value)
781
except errors.ExistingContent:
782
# No content change against a carry_over parent
783
# Perhaps this should also yield a fs hash update?
785
entry.text_size = parent_entry.text_size
786
entry.text_sha1 = parent_entry.text_sha1
787
elif kind == 'symlink':
789
entry.symlink_target = tree.get_symlink_target(file_id)
790
if (carry_over_possible and
791
parent_entry.symlink_target == entry.symlink_target):
794
self._add_text_to_weave(change[0], '', heads, None)
795
elif kind == 'directory':
796
if carry_over_possible:
799
# Nothing to set on the entry.
800
# XXX: split into the Root and nonRoot versions.
801
if change[1][1] != '' or self.repository.supports_rich_root():
802
self._add_text_to_weave(change[0], '', heads, None)
803
elif kind == 'tree-reference':
804
if not self.repository._format.supports_tree_reference:
805
# This isn't quite sane as an error, but we shouldn't
806
# ever see this code path in practice: tree's don't
807
# permit references when the repo doesn't support tree
809
raise errors.UnsupportedOperation(tree.add_reference,
811
reference_revision = tree.get_reference_revision(change[0])
812
entry.reference_revision = reference_revision
813
if (carry_over_possible and
814
parent_entry.reference_revision == reference_revision):
817
self._add_text_to_weave(change[0], '', heads, None)
819
raise AssertionError('unknown kind %r' % kind)
821
entry.revision = modified_rev
823
entry.revision = parent_entry.revision
826
new_path = change[1][1]
827
inv_delta.append((change[1][0], new_path, change[0], entry))
830
self.new_inventory = None
832
# This should perhaps be guarded by a check that the basis we
833
# commit against is the basis for the commit and if not do a delta
835
self._any_changes = True
837
# housekeeping root entry changes do not affect no-change commits.
838
self._require_root_change(tree)
839
self.basis_delta_revision = basis_revision_id
841
def _add_text_to_weave(self, file_id, new_text, parents, nostore_sha):
842
parent_keys = tuple([(file_id, parent) for parent in parents])
843
return self.repository.texts._add_text(
844
(file_id, self._new_revision_id), parent_keys, new_text,
845
nostore_sha=nostore_sha, random_id=self.random_revid)[0:2]
848
class RootCommitBuilder(CommitBuilder):
849
"""This commitbuilder actually records the root id"""
851
# the root entry gets versioned properly by this builder.
852
_versioned_root = True
854
def _check_root(self, ie, parent_invs, tree):
855
"""Helper for record_entry_contents.
857
:param ie: An entry being added.
858
:param parent_invs: The inventories of the parent revisions of the
860
:param tree: The tree that is being committed.
863
def _require_root_change(self, tree):
864
"""Enforce an appropriate root object change.
866
This is called once when record_iter_changes is called, if and only if
867
the root was not in the delta calculated by record_iter_changes.
869
:param tree: The tree which is being committed.
871
# versioned roots do not change unless the tree found a change.
227
raise NotImplementedError(self.record_iter_changes)
874
230
class RepositoryWriteLockResult(LogicalLockResult):
1029
319
return InterRepository._assert_same_model(self, repository)
1031
def add_inventory(self, revision_id, inv, parents):
1032
"""Add the inventory inv to the repository as revision_id.
1034
:param parents: The revision ids of the parents that revision_id
1035
is known to have and are in the repository already.
1037
:returns: The validator(which is a sha1 digest, though what is sha'd is
1038
repository format specific) of the serialized inventory.
1040
if not self.is_in_write_group():
1041
raise AssertionError("%r not in write group" % (self,))
1042
_mod_revision.check_not_reserved_id(revision_id)
1043
if not (inv.revision_id is None or inv.revision_id == revision_id):
1044
raise AssertionError(
1045
"Mismatch between inventory revision"
1046
" id and insertion revid (%r, %r)"
1047
% (inv.revision_id, revision_id))
1048
if inv.root is None:
1049
raise errors.RootMissing()
1050
return self._add_inventory_checked(revision_id, inv, parents)
1052
def _add_inventory_checked(self, revision_id, inv, parents):
1053
"""Add inv to the repository after checking the inputs.
1055
This function can be overridden to allow different inventory styles.
1057
:seealso: add_inventory, for the contract.
1059
inv_lines = self._serializer.write_inventory_to_lines(inv)
1060
return self._inventory_add_lines(revision_id, parents,
1061
inv_lines, check_content=False)
1063
def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
1064
parents, basis_inv=None, propagate_caches=False):
1065
"""Add a new inventory expressed as a delta against another revision.
1067
See the inventory developers documentation for the theory behind
1070
:param basis_revision_id: The inventory id the delta was created
1071
against. (This does not have to be a direct parent.)
1072
:param delta: The inventory delta (see Inventory.apply_delta for
1074
:param new_revision_id: The revision id that the inventory is being
1076
:param parents: The revision ids of the parents that revision_id is
1077
known to have and are in the repository already. These are supplied
1078
for repositories that depend on the inventory graph for revision
1079
graph access, as well as for those that pun ancestry with delta
1081
:param basis_inv: The basis inventory if it is already known,
1083
:param propagate_caches: If True, the caches for this inventory are
1084
copied to and updated for the result if possible.
1086
:returns: (validator, new_inv)
1087
The validator(which is a sha1 digest, though what is sha'd is
1088
repository format specific) of the serialized inventory, and the
1089
resulting inventory.
1091
if not self.is_in_write_group():
1092
raise AssertionError("%r not in write group" % (self,))
1093
_mod_revision.check_not_reserved_id(new_revision_id)
1094
basis_tree = self.revision_tree(basis_revision_id)
1095
basis_tree.lock_read()
1097
# Note that this mutates the inventory of basis_tree, which not all
1098
# inventory implementations may support: A better idiom would be to
1099
# return a new inventory, but as there is no revision tree cache in
1100
# repository this is safe for now - RBC 20081013
1101
if basis_inv is None:
1102
basis_inv = basis_tree.inventory
1103
basis_inv.apply_delta(delta)
1104
basis_inv.revision_id = new_revision_id
1105
return (self.add_inventory(new_revision_id, basis_inv, parents),
1110
def _inventory_add_lines(self, revision_id, parents, lines,
1111
check_content=True):
1112
"""Store lines in inv_vf and return the sha1 of the inventory."""
1113
parents = [(parent,) for parent in parents]
1114
result = self.inventories.add_lines((revision_id,), parents, lines,
1115
check_content=check_content)[0]
1116
self.inventories._access.flush()
1119
def add_revision(self, revision_id, rev, inv=None, config=None):
1120
"""Add rev to the revision store as revision_id.
1122
:param revision_id: the revision id to use.
1123
:param rev: The revision object.
1124
:param inv: The inventory for the revision. if None, it will be looked
1125
up in the inventory storer
1126
:param config: If None no digital signature will be created.
1127
If supplied its signature_needed method will be used
1128
to determine if a signature should be made.
1130
# TODO: jam 20070210 Shouldn't we check rev.revision_id and
1132
_mod_revision.check_not_reserved_id(revision_id)
1133
if config is not None and config.signature_needed():
1135
inv = self.get_inventory(revision_id)
1136
plaintext = Testament(rev, inv).as_short_text()
1137
self.store_revision_signature(
1138
gpg.GPGStrategy(config), plaintext, revision_id)
1139
# check inventory present
1140
if not self.inventories.get_parent_map([(revision_id,)]):
1142
raise errors.WeaveRevisionNotPresent(revision_id,
1145
# yes, this is not suitable for adding with ghosts.
1146
rev.inventory_sha1 = self.add_inventory(revision_id, inv,
1149
key = (revision_id,)
1150
rev.inventory_sha1 = self.inventories.get_sha1s([key])[key]
1151
self._add_revision(rev)
1153
def _add_revision(self, revision):
1154
text = self._serializer.write_revision_to_string(revision)
1155
key = (revision.revision_id,)
1156
parents = tuple((parent,) for parent in revision.parent_ids)
1157
self.revisions.add_lines(key, parents, osutils.split_lines(text))
1159
321
def all_revision_ids(self):
1160
322
"""Returns a list of all the revision ids in the repository.
1185
347
self.control_files.break_lock()
1188
def _eliminate_revisions_not_present(self, revision_ids):
1189
"""Check every revision id in revision_ids to see if we have it.
1191
Returns a set of the present revisions.
1194
graph = self.get_graph()
1195
parent_map = graph.get_parent_map(revision_ids)
1196
# The old API returned a list, should this actually be a set?
1197
return parent_map.keys()
1199
def _check_inventories(self, checker):
1200
"""Check the inventories found from the revision scan.
1202
This is responsible for verifying the sha1 of inventories and
1203
creating a pending_keys set that covers data referenced by inventories.
1205
bar = ui.ui_factory.nested_progress_bar()
1207
self._do_check_inventories(checker, bar)
1211
def _do_check_inventories(self, checker, bar):
1212
"""Helper for _check_inventories."""
1214
keys = {'chk_bytes':set(), 'inventories':set(), 'texts':set()}
1215
kinds = ['chk_bytes', 'texts']
1216
count = len(checker.pending_keys)
1217
bar.update("inventories", 0, 2)
1218
current_keys = checker.pending_keys
1219
checker.pending_keys = {}
1220
# Accumulate current checks.
1221
for key in current_keys:
1222
if key[0] != 'inventories' and key[0] not in kinds:
1223
checker._report_items.append('unknown key type %r' % (key,))
1224
keys[key[0]].add(key[1:])
1225
if keys['inventories']:
1226
# NB: output order *should* be roughly sorted - topo or
1227
# inverse topo depending on repository - either way decent
1228
# to just delta against. However, pre-CHK formats didn't
1229
# try to optimise inventory layout on disk. As such the
1230
# pre-CHK code path does not use inventory deltas.
1232
for record in self.inventories.check(keys=keys['inventories']):
1233
if record.storage_kind == 'absent':
1234
checker._report_items.append(
1235
'Missing inventory {%s}' % (record.key,))
1237
last_object = self._check_record('inventories', record,
1238
checker, last_object,
1239
current_keys[('inventories',) + record.key])
1240
del keys['inventories']
1243
bar.update("texts", 1)
1244
while (checker.pending_keys or keys['chk_bytes']
1246
# Something to check.
1247
current_keys = checker.pending_keys
1248
checker.pending_keys = {}
1249
# Accumulate current checks.
1250
for key in current_keys:
1251
if key[0] not in kinds:
1252
checker._report_items.append('unknown key type %r' % (key,))
1253
keys[key[0]].add(key[1:])
1254
# Check the outermost kind only - inventories || chk_bytes || texts
1258
for record in getattr(self, kind).check(keys=keys[kind]):
1259
if record.storage_kind == 'absent':
1260
checker._report_items.append(
1261
'Missing %s {%s}' % (kind, record.key,))
1263
last_object = self._check_record(kind, record,
1264
checker, last_object, current_keys[(kind,) + record.key])
1268
def _check_record(self, kind, record, checker, last_object, item_data):
1269
"""Check a single text from this repository."""
1270
if kind == 'inventories':
1271
rev_id = record.key[0]
1272
inv = self._deserialise_inventory(rev_id,
1273
record.get_bytes_as('fulltext'))
1274
if last_object is not None:
1275
delta = inv._make_delta(last_object)
1276
for old_path, path, file_id, ie in delta:
1279
ie.check(checker, rev_id, inv)
1281
for path, ie in inv.iter_entries():
1282
ie.check(checker, rev_id, inv)
1283
if self._format.fast_deltas:
1285
elif kind == 'chk_bytes':
1286
# No code written to check chk_bytes for this repo format.
1287
checker._report_items.append(
1288
'unsupported key type chk_bytes for %s' % (record.key,))
1289
elif kind == 'texts':
1290
self._check_text(record, checker, item_data)
1292
checker._report_items.append(
1293
'unknown key type %s for %s' % (kind, record.key))
1295
def _check_text(self, record, checker, item_data):
1296
"""Check a single text."""
1297
# Check it is extractable.
1298
# TODO: check length.
1299
if record.storage_kind == 'chunked':
1300
chunks = record.get_bytes_as(record.storage_kind)
1301
sha1 = osutils.sha_strings(chunks)
1302
length = sum(map(len, chunks))
1304
content = record.get_bytes_as('fulltext')
1305
sha1 = osutils.sha_string(content)
1306
length = len(content)
1307
if item_data and sha1 != item_data[1]:
1308
checker._report_items.append(
1309
'sha1 mismatch: %s has sha1 %s expected %s referenced by %s' %
1310
(record.key, sha1, item_data[1], item_data[2]))
1313
def create(a_bzrdir):
1314
"""Construct the current default format repository in a_bzrdir."""
1315
return RepositoryFormat.get_default_format().initialize(a_bzrdir)
350
def create(controldir):
351
"""Construct the current default format repository in controldir."""
352
return RepositoryFormat.get_default_format().initialize(controldir)
1317
def __init__(self, _format, a_bzrdir, control_files):
354
def __init__(self, _format, controldir, control_files):
1318
355
"""instantiate a Repository.
1320
357
:param _format: The format of the repository on disk.
1321
:param a_bzrdir: The BzrDir of the repository.
358
:param controldir: The ControlDir of the repository.
359
:param control_files: Control files to use for locking, etc.
1323
361
# In the future we will have a single api for all stores for
1324
362
# getting file texts, inventories and revisions, then
1997
909
signature = gpg_strategy.sign(plaintext)
1998
910
self.add_signature_text(revision_id, signature)
2001
912
def add_signature_text(self, revision_id, signature):
2002
self.signatures.add_lines((revision_id,), (),
2003
osutils.split_lines(signature))
2005
def find_text_key_references(self):
2006
"""Find the text key references within the repository.
2008
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2009
to whether they were referred to by the inventory of the
2010
revision_id that they contain. The inventory texts from all present
2011
revision ids are assessed to generate this report.
2013
revision_keys = self.revisions.keys()
2014
w = self.inventories
2015
pb = ui.ui_factory.nested_progress_bar()
2017
return self._find_text_key_references_from_xml_inventory_lines(
2018
w.iter_lines_added_or_present_in_keys(revision_keys, pb=pb))
2022
def _find_text_key_references_from_xml_inventory_lines(self,
2024
"""Core routine for extracting references to texts from inventories.
2026
This performs the translation of xml lines to revision ids.
2028
:param line_iterator: An iterator of lines, origin_version_id
2029
:return: A dictionary mapping text keys ((fileid, revision_id) tuples)
2030
to whether they were referred to by the inventory of the
2031
revision_id that they contain. Note that if that revision_id was
2032
not part of the line_iterator's output then False will be given -
2033
even though it may actually refer to that key.
2035
if not self._serializer.support_altered_by_hack:
2036
raise AssertionError(
2037
"_find_text_key_references_from_xml_inventory_lines only "
2038
"supported for branches which store inventory as unnested xml"
2039
", not on %r" % self)
2042
# this code needs to read every new line in every inventory for the
2043
# inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
2044
# not present in one of those inventories is unnecessary but not
2045
# harmful because we are filtering by the revision id marker in the
2046
# inventory lines : we only select file ids altered in one of those
2047
# revisions. We don't need to see all lines in the inventory because
2048
# only those added in an inventory in rev X can contain a revision=X
2050
unescape_revid_cache = {}
2051
unescape_fileid_cache = {}
2053
# jam 20061218 In a big fetch, this handles hundreds of thousands
2054
# of lines, so it has had a lot of inlining and optimizing done.
2055
# Sorry that it is a little bit messy.
2056
# Move several functions to be local variables, since this is a long
2058
search = self._file_ids_altered_regex.search
2059
unescape = _unescape_xml
2060
setdefault = result.setdefault
2061
for line, line_key in line_iterator:
2062
match = search(line)
2065
# One call to match.group() returning multiple items is quite a
2066
# bit faster than 2 calls to match.group() each returning 1
2067
file_id, revision_id = match.group('file_id', 'revision_id')
2069
# Inlining the cache lookups helps a lot when you make 170,000
2070
# lines and 350k ids, versus 8.4 unique ids.
2071
# Using a cache helps in 2 ways:
2072
# 1) Avoids unnecessary decoding calls
2073
# 2) Re-uses cached strings, which helps in future set and
2075
# (2) is enough that removing encoding entirely along with
2076
# the cache (so we are using plain strings) results in no
2077
# performance improvement.
2079
revision_id = unescape_revid_cache[revision_id]
2081
unescaped = unescape(revision_id)
2082
unescape_revid_cache[revision_id] = unescaped
2083
revision_id = unescaped
2085
# Note that unconditionally unescaping means that we deserialise
2086
# every fileid, which for general 'pull' is not great, but we don't
2087
# really want to have some many fulltexts that this matters anyway.
2090
file_id = unescape_fileid_cache[file_id]
2092
unescaped = unescape(file_id)
2093
unescape_fileid_cache[file_id] = unescaped
2096
key = (file_id, revision_id)
2097
setdefault(key, False)
2098
if revision_id == line_key[-1]:
2102
def _inventory_xml_lines_for_keys(self, keys):
2103
"""Get a line iterator of the sort needed for findind references.
2105
Not relevant for non-xml inventory repositories.
2107
Ghosts in revision_keys are ignored.
2109
:param revision_keys: The revision keys for the inventories to inspect.
2110
:return: An iterator over (inventory line, revid) for the fulltexts of
2111
all of the xml inventories specified by revision_keys.
2113
stream = self.inventories.get_record_stream(keys, 'unordered', True)
2114
for record in stream:
2115
if record.storage_kind != 'absent':
2116
chunks = record.get_bytes_as('chunked')
2117
revid = record.key[-1]
2118
lines = osutils.chunks_to_lines(chunks)
2122
def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
2124
"""Helper routine for fileids_altered_by_revision_ids.
2126
This performs the translation of xml lines to revision ids.
2128
:param line_iterator: An iterator of lines, origin_version_id
2129
:param revision_keys: The revision ids to filter for. This should be a
2130
set or other type which supports efficient __contains__ lookups, as
2131
the revision key from each parsed line will be looked up in the
2132
revision_keys filter.
2133
:return: a dictionary mapping altered file-ids to an iterable of
2134
revision_ids. Each altered file-ids has the exact revision_ids that
2135
altered it listed explicitly.
2137
seen = set(self._find_text_key_references_from_xml_inventory_lines(
2138
line_iterator).iterkeys())
2139
parent_keys = self._find_parent_keys_of_revisions(revision_keys)
2140
parent_seen = set(self._find_text_key_references_from_xml_inventory_lines(
2141
self._inventory_xml_lines_for_keys(parent_keys)))
2142
new_keys = seen - parent_seen
2144
setdefault = result.setdefault
2145
for key in new_keys:
2146
setdefault(key[0], set()).add(key[-1])
913
"""Store a signature text for a revision.
915
:param revision_id: Revision id of the revision
916
:param signature: Signature text.
918
raise NotImplementedError(self.add_signature_text)
2149
920
def _find_parent_ids_of_revisions(self, revision_ids):
2150
921
"""Find all parent ids that are mentioned in the revision graph.
2201
941
uniquely identify the file version in the caller's context. (Examples:
2202
942
an index number or a TreeTransform trans_id.)
2204
bytes_iterator is an iterable of bytestrings for the file. The
2205
kind of iterable and length of the bytestrings are unspecified, but for
2206
this implementation, it is a list of bytes produced by
2207
VersionedFile.get_record_stream().
2209
944
:param desired_files: a list of (file_id, revision_id, identifier)
2213
for file_id, revision_id, callable_data in desired_files:
2214
text_keys[(file_id, revision_id)] = callable_data
2215
for record in self.texts.get_record_stream(text_keys, 'unordered', True):
2216
if record.storage_kind == 'absent':
2217
raise errors.RevisionNotPresent(record.key, self)
2218
yield text_keys[record.key], record.get_bytes_as('chunked')
2220
def _generate_text_key_index(self, text_key_references=None,
2222
"""Generate a new text key index for the repository.
2224
This is an expensive function that will take considerable time to run.
2226
:return: A dict mapping text keys ((file_id, revision_id) tuples) to a
2227
list of parents, also text keys. When a given key has no parents,
2228
the parents list will be [NULL_REVISION].
2230
# All revisions, to find inventory parents.
2231
if ancestors is None:
2232
graph = self.get_graph()
2233
ancestors = graph.get_parent_map(self.all_revision_ids())
2234
if text_key_references is None:
2235
text_key_references = self.find_text_key_references()
2236
pb = ui.ui_factory.nested_progress_bar()
2238
return self._do_generate_text_key_index(ancestors,
2239
text_key_references, pb)
2243
def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
2244
"""Helper for _generate_text_key_index to avoid deep nesting."""
2245
revision_order = tsort.topo_sort(ancestors)
2246
invalid_keys = set()
2248
for revision_id in revision_order:
2249
revision_keys[revision_id] = set()
2250
text_count = len(text_key_references)
2251
# a cache of the text keys to allow reuse; costs a dict of all the
2252
# keys, but saves a 2-tuple for every child of a given key.
2254
for text_key, valid in text_key_references.iteritems():
2256
invalid_keys.add(text_key)
2258
revision_keys[text_key[1]].add(text_key)
2259
text_key_cache[text_key] = text_key
2260
del text_key_references
2262
text_graph = graph.Graph(graph.DictParentsProvider(text_index))
2263
NULL_REVISION = _mod_revision.NULL_REVISION
2264
# Set a cache with a size of 10 - this suffices for bzr.dev but may be
2265
# too small for large or very branchy trees. However, for 55K path
2266
# trees, it would be easy to use too much memory trivially. Ideally we
2267
# could gauge this by looking at available real memory etc, but this is
2268
# always a tricky proposition.
2269
inventory_cache = lru_cache.LRUCache(10)
2270
batch_size = 10 # should be ~150MB on a 55K path tree
2271
batch_count = len(revision_order) / batch_size + 1
2273
pb.update("Calculating text parents", processed_texts, text_count)
2274
for offset in xrange(batch_count):
2275
to_query = revision_order[offset * batch_size:(offset + 1) *
2279
for revision_id in to_query:
2280
parent_ids = ancestors[revision_id]
2281
for text_key in revision_keys[revision_id]:
2282
pb.update("Calculating text parents", processed_texts)
2283
processed_texts += 1
2284
candidate_parents = []
2285
for parent_id in parent_ids:
2286
parent_text_key = (text_key[0], parent_id)
2288
check_parent = parent_text_key not in \
2289
revision_keys[parent_id]
2291
# the parent parent_id is a ghost:
2292
check_parent = False
2293
# truncate the derived graph against this ghost.
2294
parent_text_key = None
2296
# look at the parent commit details inventories to
2297
# determine possible candidates in the per file graph.
2300
inv = inventory_cache[parent_id]
2302
inv = self.revision_tree(parent_id).inventory
2303
inventory_cache[parent_id] = inv
2305
parent_entry = inv[text_key[0]]
2306
except (KeyError, errors.NoSuchId):
2308
if parent_entry is not None:
2310
text_key[0], parent_entry.revision)
2312
parent_text_key = None
2313
if parent_text_key is not None:
2314
candidate_parents.append(
2315
text_key_cache[parent_text_key])
2316
parent_heads = text_graph.heads(candidate_parents)
2317
new_parents = list(parent_heads)
2318
new_parents.sort(key=lambda x:candidate_parents.index(x))
2319
if new_parents == []:
2320
new_parents = [NULL_REVISION]
2321
text_index[text_key] = new_parents
2323
for text_key in invalid_keys:
2324
text_index[text_key] = [NULL_REVISION]
2327
def item_keys_introduced_by(self, revision_ids, _files_pb=None):
2328
"""Get an iterable listing the keys of all the data introduced by a set
2331
The keys will be ordered so that the corresponding items can be safely
2332
fetched and inserted in that order.
2334
:returns: An iterable producing tuples of (knit-kind, file-id,
2335
versions). knit-kind is one of 'file', 'inventory', 'signatures',
2336
'revisions'. file-id is None unless knit-kind is 'file'.
2338
for result in self._find_file_keys_to_fetch(revision_ids, _files_pb):
2341
for result in self._find_non_file_keys_to_fetch(revision_ids):
2344
def _find_file_keys_to_fetch(self, revision_ids, pb):
2345
# XXX: it's a bit weird to control the inventory weave caching in this
2346
# generator. Ideally the caching would be done in fetch.py I think. Or
2347
# maybe this generator should explicitly have the contract that it
2348
# should not be iterated until the previously yielded item has been
2350
inv_w = self.inventories
2352
# file ids that changed
2353
file_ids = self.fileids_altered_by_revision_ids(revision_ids, inv_w)
2355
num_file_ids = len(file_ids)
2356
for file_id, altered_versions in file_ids.iteritems():
2358
pb.update("Fetch texts", count, num_file_ids)
2360
yield ("file", file_id, altered_versions)
2362
def _find_non_file_keys_to_fetch(self, revision_ids):
2364
yield ("inventory", None, revision_ids)
2367
# XXX: Note ATM no callers actually pay attention to this return
2368
# instead they just use the list of revision ids and ignore
2369
# missing sigs. Consider removing this work entirely
2370
revisions_with_signatures = set(self.signatures.get_parent_map(
2371
[(r,) for r in revision_ids]))
2372
revisions_with_signatures = set(
2373
[r for (r,) in revisions_with_signatures])
2374
revisions_with_signatures.intersection_update(revision_ids)
2375
yield ("signatures", None, revisions_with_signatures)
2378
yield ("revisions", None, revision_ids)
2381
def get_inventory(self, revision_id):
2382
"""Get Inventory object by revision id."""
2383
return self.iter_inventories([revision_id]).next()
2385
def iter_inventories(self, revision_ids, ordering=None):
2386
"""Get many inventories by revision_ids.
2388
This will buffer some or all of the texts used in constructing the
2389
inventories in memory, but will only parse a single inventory at a
2392
:param revision_ids: The expected revision ids of the inventories.
2393
:param ordering: optional ordering, e.g. 'topological'. If not
2394
specified, the order of revision_ids will be preserved (by
2395
buffering if necessary).
2396
:return: An iterator of inventories.
2398
if ((None in revision_ids)
2399
or (_mod_revision.NULL_REVISION in revision_ids)):
2400
raise ValueError('cannot get null revision inventory')
2401
return self._iter_inventories(revision_ids, ordering)
2403
def _iter_inventories(self, revision_ids, ordering):
2404
"""single-document based inventory iteration."""
2405
inv_xmls = self._iter_inventory_xmls(revision_ids, ordering)
2406
for text, revision_id in inv_xmls:
2407
yield self._deserialise_inventory(revision_id, text)
2409
def _iter_inventory_xmls(self, revision_ids, ordering):
2410
if ordering is None:
2411
order_as_requested = True
2412
ordering = 'unordered'
2414
order_as_requested = False
2415
keys = [(revision_id,) for revision_id in revision_ids]
2418
if order_as_requested:
2419
key_iter = iter(keys)
2420
next_key = key_iter.next()
2421
stream = self.inventories.get_record_stream(keys, ordering, True)
2423
for record in stream:
2424
if record.storage_kind != 'absent':
2425
chunks = record.get_bytes_as('chunked')
2426
if order_as_requested:
2427
text_chunks[record.key] = chunks
2429
yield ''.join(chunks), record.key[-1]
2431
raise errors.NoSuchRevision(self, record.key)
2432
if order_as_requested:
2433
# Yield as many results as we can while preserving order.
2434
while next_key in text_chunks:
2435
chunks = text_chunks.pop(next_key)
2436
yield ''.join(chunks), next_key[-1]
2438
next_key = key_iter.next()
2439
except StopIteration:
2440
# We still want to fully consume the get_record_stream,
2441
# just in case it is not actually finished at this point
2445
def _deserialise_inventory(self, revision_id, xml):
2446
"""Transform the xml into an inventory object.
2448
:param revision_id: The expected revision id of the inventory.
2449
:param xml: A serialised inventory.
2451
result = self._serializer.read_inventory_from_string(xml, revision_id,
2452
entry_cache=self._inventory_entry_cache,
2453
return_from_cache=self._safe_to_return_from_cache)
2454
if result.revision_id != revision_id:
2455
raise AssertionError('revision id mismatch %s != %s' % (
2456
result.revision_id, revision_id))
2459
def get_serializer_format(self):
2460
return self._serializer.format_num
2463
def _get_inventory_xml(self, revision_id):
2464
"""Get serialized inventory as a string."""
2465
texts = self._iter_inventory_xmls([revision_id], 'unordered')
2467
text, revision_id = texts.next()
2468
except StopIteration:
2469
raise errors.HistoryMissing(self, 'inventory', revision_id)
947
raise NotImplementedError(self.iter_files_bytes)
2472
949
def get_rev_id_for_revno(self, revno, known_pair):
2473
950
"""Return the revision id of a revno, given a later (revno, revid)
2806
1221
except UnicodeDecodeError:
2807
1222
raise errors.NonAsciiRevisionId(method, self)
2809
def revision_graph_can_have_wrong_parents(self):
2810
"""Is it possible for this repository to have a revision graph with
2813
If True, then this repository must also implement
2814
_find_inconsistent_revision_parents so that check and reconcile can
2815
check for inconsistencies before proceeding with other checks that may
2816
depend on the revision index being consistent.
2818
raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
2821
# remove these delegates a while after bzr 0.15
2822
def __make_delegated(name, from_module):
2823
def _deprecated_repository_forwarder():
2824
symbol_versioning.warn('%s moved to %s in bzr 0.15'
2825
% (name, from_module),
2828
m = __import__(from_module, globals(), locals(), [name])
2830
return getattr(m, name)
2831
except AttributeError:
2832
raise AttributeError('module %s has no name %s'
2834
globals()[name] = _deprecated_repository_forwarder
2837
'AllInOneRepository',
2838
'WeaveMetaDirRepository',
2839
'PreSplitOutRepositoryFormat',
2840
'RepositoryFormat4',
2841
'RepositoryFormat5',
2842
'RepositoryFormat6',
2843
'RepositoryFormat7',
2845
__make_delegated(_name, 'bzrlib.repofmt.weaverepo')
2849
'RepositoryFormatKnit',
2850
'RepositoryFormatKnit1',
2852
__make_delegated(_name, 'bzrlib.repofmt.knitrepo')
2855
def install_revision(repository, rev, revision_tree):
2856
"""Install all revision data into a repository."""
2857
install_revisions(repository, [(rev, revision_tree, None)])
2860
def install_revisions(repository, iterable, num_revisions=None, pb=None):
2861
"""Install all revision data into a repository.
2863
Accepts an iterable of revision, tree, signature tuples. The signature
2866
repository.start_write_group()
2868
inventory_cache = lru_cache.LRUCache(10)
2869
for n, (revision, revision_tree, signature) in enumerate(iterable):
2870
_install_revision(repository, revision, revision_tree, signature,
2873
pb.update('Transferring revisions', n + 1, num_revisions)
2875
repository.abort_write_group()
2878
repository.commit_write_group()
2881
def _install_revision(repository, rev, revision_tree, signature,
2883
"""Install all revision data into a repository."""
2884
present_parents = []
2886
for p_id in rev.parent_ids:
2887
if repository.has_revision(p_id):
2888
present_parents.append(p_id)
2889
parent_trees[p_id] = repository.revision_tree(p_id)
2891
parent_trees[p_id] = repository.revision_tree(
2892
_mod_revision.NULL_REVISION)
2894
inv = revision_tree.inventory
2895
entries = inv.iter_entries()
2896
# backwards compatibility hack: skip the root id.
2897
if not repository.supports_rich_root():
2898
path, root = entries.next()
2899
if root.revision != rev.revision_id:
2900
raise errors.IncompatibleRevision(repr(repository))
2902
for path, ie in entries:
2903
text_keys[(ie.file_id, ie.revision)] = ie
2904
text_parent_map = repository.texts.get_parent_map(text_keys)
2905
missing_texts = set(text_keys) - set(text_parent_map)
2906
# Add the texts that are not already present
2907
for text_key in missing_texts:
2908
ie = text_keys[text_key]
2910
# FIXME: TODO: The following loop overlaps/duplicates that done by
2911
# commit to determine parents. There is a latent/real bug here where
2912
# the parents inserted are not those commit would do - in particular
2913
# they are not filtered by heads(). RBC, AB
2914
for revision, tree in parent_trees.iteritems():
2915
if ie.file_id not in tree:
2917
parent_id = tree.inventory[ie.file_id].revision
2918
if parent_id in text_parents:
2920
text_parents.append((ie.file_id, parent_id))
2921
lines = revision_tree.get_file(ie.file_id).readlines()
2922
repository.texts.add_lines(text_key, text_parents, lines)
2924
# install the inventory
2925
if repository._format._commit_inv_deltas and len(rev.parent_ids):
2926
# Cache this inventory
2927
inventory_cache[rev.revision_id] = inv
2929
basis_inv = inventory_cache[rev.parent_ids[0]]
2931
repository.add_inventory(rev.revision_id, inv, present_parents)
2933
delta = inv._make_delta(basis_inv)
2934
repository.add_inventory_by_delta(rev.parent_ids[0], delta,
2935
rev.revision_id, present_parents)
2937
repository.add_inventory(rev.revision_id, inv, present_parents)
2938
except errors.RevisionAlreadyPresent:
2940
if signature is not None:
2941
repository.add_signature_text(rev.revision_id, signature)
2942
repository.add_revision(rev.revision_id, rev, inv)
2945
1225
class MetaDirRepository(Repository):
2946
1226
"""Repositories in the new meta-dir layout.
3311
1558
# NOTE: These are experimental in 0.92. Stable in 1.0 and above
3312
1559
format_registry.register_lazy(
3313
1560
'Bazaar pack repository format 1 (needs bzr 0.92)\n',
3314
'bzrlib.repofmt.pack_repo',
1561
'bzrlib.repofmt.knitpack_repo',
3315
1562
'RepositoryFormatKnitPack1',
3317
1564
format_registry.register_lazy(
3318
1565
'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
3319
'bzrlib.repofmt.pack_repo',
1566
'bzrlib.repofmt.knitpack_repo',
3320
1567
'RepositoryFormatKnitPack3',
3322
1569
format_registry.register_lazy(
3323
1570
'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
3324
'bzrlib.repofmt.pack_repo',
1571
'bzrlib.repofmt.knitpack_repo',
3325
1572
'RepositoryFormatKnitPack4',
3327
1574
format_registry.register_lazy(
3328
1575
'Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n',
3329
'bzrlib.repofmt.pack_repo',
1576
'bzrlib.repofmt.knitpack_repo',
3330
1577
'RepositoryFormatKnitPack5',
3332
1579
format_registry.register_lazy(
3333
1580
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n',
3334
'bzrlib.repofmt.pack_repo',
1581
'bzrlib.repofmt.knitpack_repo',
3335
1582
'RepositoryFormatKnitPack5RichRoot',
3337
1584
format_registry.register_lazy(
3338
1585
'Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n',
3339
'bzrlib.repofmt.pack_repo',
1586
'bzrlib.repofmt.knitpack_repo',
3340
1587
'RepositoryFormatKnitPack5RichRootBroken',
3342
1589
format_registry.register_lazy(
3343
1590
'Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n',
3344
'bzrlib.repofmt.pack_repo',
1591
'bzrlib.repofmt.knitpack_repo',
3345
1592
'RepositoryFormatKnitPack6',
3347
1594
format_registry.register_lazy(
3348
1595
'Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n',
3349
'bzrlib.repofmt.pack_repo',
1596
'bzrlib.repofmt.knitpack_repo',
3350
1597
'RepositoryFormatKnitPack6RichRoot',
1599
format_registry.register_lazy(
1600
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
1601
'bzrlib.repofmt.groupcompress_repo',
1602
'RepositoryFormat2a',
3353
1605
# Development formats.
3354
# Obsolete but kept pending a CHK based subtree format.
1606
# Check their docstrings to see if/when they are obsolete.
3355
1607
format_registry.register_lazy(
3356
1608
("Bazaar development format 2 with subtree support "
3357
1609
"(needs bzr.dev from before 1.8)\n"),
3358
'bzrlib.repofmt.pack_repo',
1610
'bzrlib.repofmt.knitpack_repo',
3359
1611
'RepositoryFormatPackDevelopment2Subtree',
3362
# 1.14->1.16 go below here
3363
format_registry.register_lazy(
3364
'Bazaar development format - group compression and chk inventory'
3365
' (needs bzr.dev from 1.14)\n',
3366
'bzrlib.repofmt.groupcompress_repo',
3367
'RepositoryFormatCHK1',
3370
format_registry.register_lazy(
3371
'Bazaar development format - chk repository with bencode revision '
3372
'serialization (needs bzr.dev from 1.16)\n',
3373
'bzrlib.repofmt.groupcompress_repo',
3374
'RepositoryFormatCHK2',
3376
format_registry.register_lazy(
3377
'Bazaar repository format 2a (needs bzr 1.16 or later)\n',
3378
'bzrlib.repofmt.groupcompress_repo',
3379
'RepositoryFormat2a',
3381
1613
format_registry.register_lazy(
3382
1614
'Bazaar development format 8\n',
3383
1615
'bzrlib.repofmt.groupcompress_repo',
3418
1649
self.target.fetch(self.source, revision_id=revision_id)
3420
1651
@needs_write_lock
3421
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
1652
def fetch(self, revision_id=None, find_ghosts=False):
3423
1653
"""Fetch the content required to construct revision_id.
3425
1655
The content is copied from self.source to self.target.
3427
1657
:param revision_id: if None all content is copied, if NULL_REVISION no
3428
1658
content is copied.
3432
ui.ui_factory.warn_experimental_format_fetch(self)
3433
from bzrlib.fetch import RepoFetcher
3434
# See <https://launchpad.net/bugs/456077> asking for a warning here
3435
if self.source._format.network_name() != self.target._format.network_name():
3436
ui.ui_factory.show_user_warning('cross_format_fetch',
3437
from_format=self.source._format,
3438
to_format=self.target._format)
3439
f = RepoFetcher(to_repository=self.target,
3440
from_repository=self.source,
3441
last_revision=revision_id,
3442
fetch_spec=fetch_spec,
3443
find_ghosts=find_ghosts)
3445
def _walk_to_common_revisions(self, revision_ids):
3446
"""Walk out from revision_ids in source to revisions target has.
3448
:param revision_ids: The start point for the search.
3449
:return: A set of revision ids.
3451
target_graph = self.target.get_graph()
3452
revision_ids = frozenset(revision_ids)
3453
missing_revs = set()
3454
source_graph = self.source.get_graph()
3455
# ensure we don't pay silly lookup costs.
3456
searcher = source_graph._make_breadth_first_searcher(revision_ids)
3457
null_set = frozenset([_mod_revision.NULL_REVISION])
3458
searcher_exhausted = False
3462
# Iterate the searcher until we have enough next_revs
3463
while len(next_revs) < self._walk_to_common_revisions_batch_size:
3465
next_revs_part, ghosts_part = searcher.next_with_ghosts()
3466
next_revs.update(next_revs_part)
3467
ghosts.update(ghosts_part)
3468
except StopIteration:
3469
searcher_exhausted = True
3471
# If there are ghosts in the source graph, and the caller asked for
3472
# them, make sure that they are present in the target.
3473
# We don't care about other ghosts as we can't fetch them and
3474
# haven't been asked to.
3475
ghosts_to_check = set(revision_ids.intersection(ghosts))
3476
revs_to_get = set(next_revs).union(ghosts_to_check)
3478
have_revs = set(target_graph.get_parent_map(revs_to_get))
3479
# we always have NULL_REVISION present.
3480
have_revs = have_revs.union(null_set)
3481
# Check if the target is missing any ghosts we need.
3482
ghosts_to_check.difference_update(have_revs)
3484
# One of the caller's revision_ids is a ghost in both the
3485
# source and the target.
3486
raise errors.NoSuchRevision(
3487
self.source, ghosts_to_check.pop())
3488
missing_revs.update(next_revs - have_revs)
3489
# Because we may have walked past the original stop point, make
3490
# sure everything is stopped
3491
stop_revs = searcher.find_seen_ancestors(have_revs)
3492
searcher.stop_searching_any(stop_revs)
3493
if searcher_exhausted:
3495
return searcher.get_result()
1661
raise NotImplementedError(self.fetch)
3497
1663
@needs_read_lock
3498
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
1664
def search_missing_revision_ids(self,
1665
revision_id=symbol_versioning.DEPRECATED_PARAMETER,
1666
find_ghosts=True, revision_ids=None, if_present_ids=None,
3499
1668
"""Return the revision ids that source has that target does not.
3501
1670
:param revision_id: only return revision ids included by this
1672
:param revision_ids: return revision ids included by these
1673
revision_ids. NoSuchRevision will be raised if any of these
1674
revisions are not present.
1675
:param if_present_ids: like revision_ids, but will not cause
1676
NoSuchRevision if any of these are absent, instead they will simply
1677
not be in the result. This is useful for e.g. finding revisions
1678
to fetch for tags, which may reference absent revisions.
3503
1679
:param find_ghosts: If True find missing revisions in deep history
3504
1680
rather than just finding the surface difference.
1681
:param limit: Maximum number of revisions to return, topologically
3505
1683
:return: A bzrlib.graph.SearchResult.
3507
# stop searching at found target revisions.
3508
if not find_ghosts and revision_id is not None:
3509
return self._walk_to_common_revisions([revision_id])
3510
# generic, possibly worst case, slow code path.
3511
target_ids = set(self.target.all_revision_ids())
3512
if revision_id is not None:
3513
source_ids = self.source.get_ancestry(revision_id)
3514
if source_ids[0] is not None:
3515
raise AssertionError()
3518
source_ids = self.source.all_revision_ids()
3519
result_set = set(source_ids).difference(target_ids)
3520
return self.source.revision_ids_to_search_result(result_set)
1685
raise NotImplementedError(self.search_missing_revision_ids)
3523
1688
def _same_model(source, target):
3544
1709
"different serializers")
3547
class InterSameDataRepository(InterRepository):
3548
"""Code for converting between repositories that represent the same data.
3550
Data format and model must match for this to work.
3554
def _get_repo_format_to_test(self):
3555
"""Repository format for testing with.
3557
InterSameData can pull from subtree to subtree and from non-subtree to
3558
non-subtree, so we test this with the richest repository format.
3560
from bzrlib.repofmt import knitrepo
3561
return knitrepo.RepositoryFormatKnit3()
3564
def is_compatible(source, target):
3565
return InterRepository._same_model(source, target)
3568
class InterWeaveRepo(InterSameDataRepository):
3569
"""Optimised code paths between Weave based repositories.
3571
This should be in bzrlib/repofmt/weaverepo.py but we have not yet
3572
implemented lazy inter-object optimisation.
3576
def _get_repo_format_to_test(self):
3577
from bzrlib.repofmt import weaverepo
3578
return weaverepo.RepositoryFormat7()
3581
def is_compatible(source, target):
3582
"""Be compatible with known Weave formats.
3584
We don't test for the stores being of specific types because that
3585
could lead to confusing results, and there is no need to be
3588
from bzrlib.repofmt.weaverepo import (
3594
return (isinstance(source._format, (RepositoryFormat5,
3596
RepositoryFormat7)) and
3597
isinstance(target._format, (RepositoryFormat5,
3599
RepositoryFormat7)))
3600
except AttributeError:
3604
def copy_content(self, revision_id=None):
3605
"""See InterRepository.copy_content()."""
3606
# weave specific optimised path:
3608
self.target.set_make_working_trees(self.source.make_working_trees())
3609
except (errors.RepositoryUpgradeRequired, NotImplemented):
3611
# FIXME do not peek!
3612
if self.source._transport.listable():
3613
pb = ui.ui_factory.nested_progress_bar()
3615
self.target.texts.insert_record_stream(
3616
self.source.texts.get_record_stream(
3617
self.source.texts.keys(), 'topological', False))
3618
pb.update('Copying inventory', 0, 1)
3619
self.target.inventories.insert_record_stream(
3620
self.source.inventories.get_record_stream(
3621
self.source.inventories.keys(), 'topological', False))
3622
self.target.signatures.insert_record_stream(
3623
self.source.signatures.get_record_stream(
3624
self.source.signatures.keys(),
3626
self.target.revisions.insert_record_stream(
3627
self.source.revisions.get_record_stream(
3628
self.source.revisions.keys(),
3629
'topological', True))
3633
self.target.fetch(self.source, revision_id=revision_id)
3636
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3637
"""See InterRepository.missing_revision_ids()."""
3638
# we want all revisions to satisfy revision_id in source.
3639
# but we don't want to stat every file here and there.
3640
# we want then, all revisions other needs to satisfy revision_id
3641
# checked, but not those that we have locally.
3642
# so the first thing is to get a subset of the revisions to
3643
# satisfy revision_id in source, and then eliminate those that
3644
# we do already have.
3645
# this is slow on high latency connection to self, but as this
3646
# disk format scales terribly for push anyway due to rewriting
3647
# inventory.weave, this is considered acceptable.
3649
if revision_id is not None:
3650
source_ids = self.source.get_ancestry(revision_id)
3651
if source_ids[0] is not None:
3652
raise AssertionError()
3655
source_ids = self.source._all_possible_ids()
3656
source_ids_set = set(source_ids)
3657
# source_ids is the worst possible case we may need to pull.
3658
# now we want to filter source_ids against what we actually
3659
# have in target, but don't try to check for existence where we know
3660
# we do not have a revision as that would be pointless.
3661
target_ids = set(self.target._all_possible_ids())
3662
possibly_present_revisions = target_ids.intersection(source_ids_set)
3663
actually_present_revisions = set(
3664
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3665
required_revisions = source_ids_set.difference(actually_present_revisions)
3666
if revision_id is not None:
3667
# we used get_ancestry to determine source_ids then we are assured all
3668
# revisions referenced are present as they are installed in topological order.
3669
# and the tip revision was validated by get_ancestry.
3670
result_set = required_revisions
3672
# if we just grabbed the possibly available ids, then
3673
# we only have an estimate of whats available and need to validate
3674
# that against the revision records.
3676
self.source._eliminate_revisions_not_present(required_revisions))
3677
return self.source.revision_ids_to_search_result(result_set)
3680
class InterKnitRepo(InterSameDataRepository):
3681
"""Optimised code paths between Knit based repositories."""
3684
def _get_repo_format_to_test(self):
3685
from bzrlib.repofmt import knitrepo
3686
return knitrepo.RepositoryFormatKnit1()
3689
def is_compatible(source, target):
3690
"""Be compatible with known Knit formats.
3692
We don't test for the stores being of specific types because that
3693
could lead to confusing results, and there is no need to be
3696
from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
3698
are_knits = (isinstance(source._format, RepositoryFormatKnit) and
3699
isinstance(target._format, RepositoryFormatKnit))
3700
except AttributeError:
3702
return are_knits and InterRepository._same_model(source, target)
3705
def search_missing_revision_ids(self, revision_id=None, find_ghosts=True):
3706
"""See InterRepository.missing_revision_ids()."""
3707
if revision_id is not None:
3708
source_ids = self.source.get_ancestry(revision_id)
3709
if source_ids[0] is not None:
3710
raise AssertionError()
3713
source_ids = self.source.all_revision_ids()
3714
source_ids_set = set(source_ids)
3715
# source_ids is the worst possible case we may need to pull.
3716
# now we want to filter source_ids against what we actually
3717
# have in target, but don't try to check for existence where we know
3718
# we do not have a revision as that would be pointless.
3719
target_ids = set(self.target.all_revision_ids())
3720
possibly_present_revisions = target_ids.intersection(source_ids_set)
3721
actually_present_revisions = set(
3722
self.target._eliminate_revisions_not_present(possibly_present_revisions))
3723
required_revisions = source_ids_set.difference(actually_present_revisions)
3724
if revision_id is not None:
3725
# we used get_ancestry to determine source_ids then we are assured all
3726
# revisions referenced are present as they are installed in topological order.
3727
# and the tip revision was validated by get_ancestry.
3728
result_set = required_revisions
3730
# if we just grabbed the possibly available ids, then
3731
# we only have an estimate of whats available and need to validate
3732
# that against the revision records.
3734
self.source._eliminate_revisions_not_present(required_revisions))
3735
return self.source.revision_ids_to_search_result(result_set)
3738
class InterDifferingSerializer(InterRepository):
3741
def _get_repo_format_to_test(self):
3745
def is_compatible(source, target):
3746
"""Be compatible with Knit2 source and Knit3 target"""
3747
# This is redundant with format.check_conversion_target(), however that
3748
# raises an exception, and we just want to say "False" as in we won't
3749
# support converting between these formats.
3750
if 'IDS_never' in debug.debug_flags:
3752
if source.supports_rich_root() and not target.supports_rich_root():
3754
if (source._format.supports_tree_reference
3755
and not target._format.supports_tree_reference):
3757
if target._fallback_repositories and target._format.supports_chks:
3758
# IDS doesn't know how to copy CHKs for the parent inventories it
3759
# adds to stacked repos.
3761
if 'IDS_always' in debug.debug_flags:
3763
# Only use this code path for local source and target. IDS does far
3764
# too much IO (both bandwidth and roundtrips) over a network.
3765
if not source.bzrdir.transport.base.startswith('file:///'):
3767
if not target.bzrdir.transport.base.startswith('file:///'):
3771
def _get_trees(self, revision_ids, cache):
3773
for rev_id in revision_ids:
3775
possible_trees.append((rev_id, cache[rev_id]))
3777
# Not cached, but inventory might be present anyway.
3779
tree = self.source.revision_tree(rev_id)
3780
except errors.NoSuchRevision:
3781
# Nope, parent is ghost.
3784
cache[rev_id] = tree
3785
possible_trees.append((rev_id, tree))
3786
return possible_trees
3788
def _get_delta_for_revision(self, tree, parent_ids, possible_trees):
3789
"""Get the best delta and base for this revision.
3791
:return: (basis_id, delta)
3794
# Generate deltas against each tree, to find the shortest.
3795
texts_possibly_new_in_tree = set()
3796
for basis_id, basis_tree in possible_trees:
3797
delta = tree.inventory._make_delta(basis_tree.inventory)
3798
for old_path, new_path, file_id, new_entry in delta:
3799
if new_path is None:
3800
# This file_id isn't present in the new rev, so we don't
3804
# Rich roots are handled elsewhere...
3806
kind = new_entry.kind
3807
if kind != 'directory' and kind != 'file':
3808
# No text record associated with this inventory entry.
3810
# This is a directory or file that has changed somehow.
3811
texts_possibly_new_in_tree.add((file_id, new_entry.revision))
3812
deltas.append((len(delta), basis_id, delta))
3814
return deltas[0][1:]
3816
def _fetch_parent_invs_for_stacking(self, parent_map, cache):
3817
"""Find all parent revisions that are absent, but for which the
3818
inventory is present, and copy those inventories.
3820
This is necessary to preserve correctness when the source is stacked
3821
without fallbacks configured. (Note that in cases like upgrade the
3822
source may be not have _fallback_repositories even though it is
3826
for parents in parent_map.values():
3827
parent_revs.update(parents)
3828
present_parents = self.source.get_parent_map(parent_revs)
3829
absent_parents = set(parent_revs).difference(present_parents)
3830
parent_invs_keys_for_stacking = self.source.inventories.get_parent_map(
3831
(rev_id,) for rev_id in absent_parents)
3832
parent_inv_ids = [key[-1] for key in parent_invs_keys_for_stacking]
3833
for parent_tree in self.source.revision_trees(parent_inv_ids):
3834
current_revision_id = parent_tree.get_revision_id()
3835
parents_parents_keys = parent_invs_keys_for_stacking[
3836
(current_revision_id,)]
3837
parents_parents = [key[-1] for key in parents_parents_keys]
3838
basis_id = _mod_revision.NULL_REVISION
3839
basis_tree = self.source.revision_tree(basis_id)
3840
delta = parent_tree.inventory._make_delta(basis_tree.inventory)
3841
self.target.add_inventory_by_delta(
3842
basis_id, delta, current_revision_id, parents_parents)
3843
cache[current_revision_id] = parent_tree
3845
def _fetch_batch(self, revision_ids, basis_id, cache):
3846
"""Fetch across a few revisions.
3848
:param revision_ids: The revisions to copy
3849
:param basis_id: The revision_id of a tree that must be in cache, used
3850
as a basis for delta when no other base is available
3851
:param cache: A cache of RevisionTrees that we can use.
3852
:return: The revision_id of the last converted tree. The RevisionTree
3853
for it will be in cache
3855
# Walk though all revisions; get inventory deltas, copy referenced
3856
# texts that delta references, insert the delta, revision and
3858
root_keys_to_create = set()
3861
pending_revisions = []
3862
parent_map = self.source.get_parent_map(revision_ids)
3863
self._fetch_parent_invs_for_stacking(parent_map, cache)
3864
self.source._safe_to_return_from_cache = True
3865
for tree in self.source.revision_trees(revision_ids):
3866
# Find a inventory delta for this revision.
3867
# Find text entries that need to be copied, too.
3868
current_revision_id = tree.get_revision_id()
3869
parent_ids = parent_map.get(current_revision_id, ())
3870
parent_trees = self._get_trees(parent_ids, cache)
3871
possible_trees = list(parent_trees)
3872
if len(possible_trees) == 0:
3873
# There either aren't any parents, or the parents are ghosts,
3874
# so just use the last converted tree.
3875
possible_trees.append((basis_id, cache[basis_id]))
3876
basis_id, delta = self._get_delta_for_revision(tree, parent_ids,
3878
revision = self.source.get_revision(current_revision_id)
3879
pending_deltas.append((basis_id, delta,
3880
current_revision_id, revision.parent_ids))
3881
if self._converting_to_rich_root:
3882
self._revision_id_to_root_id[current_revision_id] = \
3884
# Determine which texts are in present in this revision but not in
3885
# any of the available parents.
3886
texts_possibly_new_in_tree = set()
3887
for old_path, new_path, file_id, entry in delta:
3888
if new_path is None:
3889
# This file_id isn't present in the new rev
3893
if not self.target.supports_rich_root():
3894
# The target doesn't support rich root, so we don't
3897
if self._converting_to_rich_root:
3898
# This can't be copied normally, we have to insert
3900
root_keys_to_create.add((file_id, entry.revision))
3903
texts_possibly_new_in_tree.add((file_id, entry.revision))
3904
for basis_id, basis_tree in possible_trees:
3905
basis_inv = basis_tree.inventory
3906
for file_key in list(texts_possibly_new_in_tree):
3907
file_id, file_revision = file_key
3909
entry = basis_inv[file_id]
3910
except errors.NoSuchId:
3912
if entry.revision == file_revision:
3913
texts_possibly_new_in_tree.remove(file_key)
3914
text_keys.update(texts_possibly_new_in_tree)
3915
pending_revisions.append(revision)
3916
cache[current_revision_id] = tree
3917
basis_id = current_revision_id
3918
self.source._safe_to_return_from_cache = False
3920
from_texts = self.source.texts
3921
to_texts = self.target.texts
3922
if root_keys_to_create:
3923
root_stream = _mod_fetch._new_root_data_stream(
3924
root_keys_to_create, self._revision_id_to_root_id, parent_map,
3926
to_texts.insert_record_stream(root_stream)
3927
to_texts.insert_record_stream(from_texts.get_record_stream(
3928
text_keys, self.target._format._fetch_order,
3929
not self.target._format._fetch_uses_deltas))
3930
# insert inventory deltas
3931
for delta in pending_deltas:
3932
self.target.add_inventory_by_delta(*delta)
3933
if self.target._fallback_repositories:
3934
# Make sure this stacked repository has all the parent inventories
3935
# for the new revisions that we are about to insert. We do this
3936
# before adding the revisions so that no revision is added until
3937
# all the inventories it may depend on are added.
3938
# Note that this is overzealous, as we may have fetched these in an
3941
revision_ids = set()
3942
for revision in pending_revisions:
3943
revision_ids.add(revision.revision_id)
3944
parent_ids.update(revision.parent_ids)
3945
parent_ids.difference_update(revision_ids)
3946
parent_ids.discard(_mod_revision.NULL_REVISION)
3947
parent_map = self.source.get_parent_map(parent_ids)
3948
# we iterate over parent_map and not parent_ids because we don't
3949
# want to try copying any revision which is a ghost
3950
for parent_tree in self.source.revision_trees(parent_map):
3951
current_revision_id = parent_tree.get_revision_id()
3952
parents_parents = parent_map[current_revision_id]
3953
possible_trees = self._get_trees(parents_parents, cache)
3954
if len(possible_trees) == 0:
3955
# There either aren't any parents, or the parents are
3956
# ghosts, so just use the last converted tree.
3957
possible_trees.append((basis_id, cache[basis_id]))
3958
basis_id, delta = self._get_delta_for_revision(parent_tree,
3959
parents_parents, possible_trees)
3960
self.target.add_inventory_by_delta(
3961
basis_id, delta, current_revision_id, parents_parents)
3962
# insert signatures and revisions
3963
for revision in pending_revisions:
3965
signature = self.source.get_signature_text(
3966
revision.revision_id)
3967
self.target.add_signature_text(revision.revision_id,
3969
except errors.NoSuchRevision:
3971
self.target.add_revision(revision.revision_id, revision)
3974
def _fetch_all_revisions(self, revision_ids, pb):
3975
"""Fetch everything for the list of revisions.
3977
:param revision_ids: The list of revisions to fetch. Must be in
3979
:param pb: A ProgressTask
3982
basis_id, basis_tree = self._get_basis(revision_ids[0])
3984
cache = lru_cache.LRUCache(100)
3985
cache[basis_id] = basis_tree
3986
del basis_tree # We don't want to hang on to it here
3990
for offset in range(0, len(revision_ids), batch_size):
3991
self.target.start_write_group()
3993
pb.update('Transferring revisions', offset,
3995
batch = revision_ids[offset:offset+batch_size]
3996
basis_id = self._fetch_batch(batch, basis_id, cache)
3998
self.source._safe_to_return_from_cache = False
3999
self.target.abort_write_group()
4002
hint = self.target.commit_write_group()
4005
if hints and self.target._format.pack_compresses:
4006
self.target.pack(hint=hints)
4007
pb.update('Transferring revisions', len(revision_ids),
4011
def fetch(self, revision_id=None, pb=None, find_ghosts=False,
4013
"""See InterRepository.fetch()."""
4014
if fetch_spec is not None:
4015
raise AssertionError("Not implemented yet...")
4016
ui.ui_factory.warn_experimental_format_fetch(self)
4017
if (not self.source.supports_rich_root()
4018
and self.target.supports_rich_root()):
4019
self._converting_to_rich_root = True
4020
self._revision_id_to_root_id = {}
4022
self._converting_to_rich_root = False
4023
# See <https://launchpad.net/bugs/456077> asking for a warning here
4024
if self.source._format.network_name() != self.target._format.network_name():
4025
ui.ui_factory.show_user_warning('cross_format_fetch',
4026
from_format=self.source._format,
4027
to_format=self.target._format)
4028
revision_ids = self.target.search_missing_revision_ids(self.source,
4029
revision_id, find_ghosts=find_ghosts).get_keys()
4030
if not revision_ids:
4032
revision_ids = tsort.topo_sort(
4033
self.source.get_graph().get_parent_map(revision_ids))
4034
if not revision_ids:
4036
# Walk though all revisions; get inventory deltas, copy referenced
4037
# texts that delta references, insert the delta, revision and
4040
my_pb = ui.ui_factory.nested_progress_bar()
4043
symbol_versioning.warn(
4044
symbol_versioning.deprecated_in((1, 14, 0))
4045
% "pb parameter to fetch()")
4048
self._fetch_all_revisions(revision_ids, pb)
4050
if my_pb is not None:
4052
return len(revision_ids), 0
4054
def _get_basis(self, first_revision_id):
4055
"""Get a revision and tree which exists in the target.
4057
This assumes that first_revision_id is selected for transmission
4058
because all other ancestors are already present. If we can't find an
4059
ancestor we fall back to NULL_REVISION since we know that is safe.
4061
:return: (basis_id, basis_tree)
4063
first_rev = self.source.get_revision(first_revision_id)
4065
basis_id = first_rev.parent_ids[0]
4066
# only valid as a basis if the target has it
4067
self.target.get_revision(basis_id)
4068
# Try to get a basis tree - if it's a ghost it will hit the
4069
# NoSuchRevision case.
4070
basis_tree = self.source.revision_tree(basis_id)
4071
except (IndexError, errors.NoSuchRevision):
4072
basis_id = _mod_revision.NULL_REVISION
4073
basis_tree = self.source.revision_tree(basis_id)
4074
return basis_id, basis_tree
4077
InterRepository.register_optimiser(InterDifferingSerializer)
4078
InterRepository.register_optimiser(InterSameDataRepository)
4079
InterRepository.register_optimiser(InterWeaveRepo)
4080
InterRepository.register_optimiser(InterKnitRepo)
4083
1712
class CopyConverter(object):
4084
1713
"""A repository conversion tool which just performs a copy of the content.
4106
1735
# trigger an assertion if not such
4107
1736
repo._format.get_format_string()
4108
1737
self.repo_dir = repo.bzrdir
4109
pb.update('Moving repository to repository.backup')
1738
pb.update(gettext('Moving repository to repository.backup'))
4110
1739
self.repo_dir.transport.move('repository', 'repository.backup')
4111
1740
backup_transport = self.repo_dir.transport.clone('repository.backup')
4112
1741
repo._format.check_conversion_target(self.target_format)
4113
1742
self.source_repo = repo._format.open(self.repo_dir,
4115
1744
_override_transport=backup_transport)
4116
pb.update('Creating new repository')
1745
pb.update(gettext('Creating new repository'))
4117
1746
converted = self.target_format.initialize(self.repo_dir,
4118
1747
self.source_repo.is_shared())
4119
1748
converted.lock_write()
4121
pb.update('Copying content')
1750
pb.update(gettext('Copying content'))
4122
1751
self.source_repo.copy_content_into(converted)
4124
1753
converted.unlock()
4125
pb.update('Deleting old repository content')
1754
pb.update(gettext('Deleting old repository content'))
4126
1755
self.repo_dir.transport.delete_tree('repository.backup')
4127
ui.ui_factory.note('repository converted')
1756
ui.ui_factory.note(gettext('repository converted'))
4140
def _unescaper(match, _map=_unescape_map):
4141
code = match.group(1)
4145
if not code.startswith('#'):
4147
return unichr(int(code[1:])).encode('utf8')
4153
def _unescape_xml(data):
4154
"""Unescape predefined XML entities in a string of data."""
4156
if _unescape_re is None:
4157
_unescape_re = re.compile('\&([^;]*);')
4158
return _unescape_re.sub(_unescaper, data)
4161
class _VersionedFileChecker(object):
4163
def __init__(self, repository, text_key_references=None, ancestors=None):
4164
self.repository = repository
4165
self.text_index = self.repository._generate_text_key_index(
4166
text_key_references=text_key_references, ancestors=ancestors)
4168
def calculate_file_version_parents(self, text_key):
4169
"""Calculate the correct parents for a file version according to
4172
parent_keys = self.text_index[text_key]
4173
if parent_keys == [_mod_revision.NULL_REVISION]:
4175
return tuple(parent_keys)
4177
def check_file_version_parents(self, texts, progress_bar=None):
4178
"""Check the parents stored in a versioned file are correct.
4180
It also detects file versions that are not referenced by their
4181
corresponding revision's inventory.
4183
:returns: A tuple of (wrong_parents, dangling_file_versions).
4184
wrong_parents is a dict mapping {revision_id: (stored_parents,
4185
correct_parents)} for each revision_id where the stored parents
4186
are not correct. dangling_file_versions is a set of (file_id,
4187
revision_id) tuples for versions that are present in this versioned
4188
file, but not used by the corresponding inventory.
4190
local_progress = None
4191
if progress_bar is None:
4192
local_progress = ui.ui_factory.nested_progress_bar()
4193
progress_bar = local_progress
4195
return self._check_file_version_parents(texts, progress_bar)
4198
local_progress.finished()
4200
def _check_file_version_parents(self, texts, progress_bar):
4201
"""See check_file_version_parents."""
4203
self.file_ids = set([file_id for file_id, _ in
4204
self.text_index.iterkeys()])
4205
# text keys is now grouped by file_id
4206
n_versions = len(self.text_index)
4207
progress_bar.update('loading text store', 0, n_versions)
4208
parent_map = self.repository.texts.get_parent_map(self.text_index)
4209
# On unlistable transports this could well be empty/error...
4210
text_keys = self.repository.texts.keys()
4211
unused_keys = frozenset(text_keys) - set(self.text_index)
4212
for num, key in enumerate(self.text_index.iterkeys()):
4213
progress_bar.update('checking text graph', num, n_versions)
4214
correct_parents = self.calculate_file_version_parents(key)
4216
knit_parents = parent_map[key]
4217
except errors.RevisionNotPresent:
4220
if correct_parents != knit_parents:
4221
wrong_parents[key] = (knit_parents, correct_parents)
4222
return wrong_parents, unused_keys
4225
def _old_get_graph(repository, revision_id):
4226
"""DO NOT USE. That is all. I'm serious."""
4227
graph = repository.get_graph()
4228
revision_graph = dict(((key, value) for key, value in
4229
graph.iter_ancestry([revision_id]) if value is not None))
4230
return _strip_NULL_ghosts(revision_graph)
4233
1760
def _strip_NULL_ghosts(revision_graph):
4234
1761
"""Also don't use this. more compatibility code for unmigrated clients."""
4235
1762
# Filter ghosts, and null:
4241
1768
return revision_graph
4244
class StreamSink(object):
4245
"""An object that can insert a stream into a repository.
4247
This interface handles the complexity of reserialising inventories and
4248
revisions from different formats, and allows unidirectional insertion into
4249
stacked repositories without looking for the missing basis parents
4253
def __init__(self, target_repo):
4254
self.target_repo = target_repo
4256
def insert_stream(self, stream, src_format, resume_tokens):
4257
"""Insert a stream's content into the target repository.
4259
:param src_format: a bzr repository format.
4261
:return: a list of resume tokens and an iterable of keys additional
4262
items required before the insertion can be completed.
4264
self.target_repo.lock_write()
4267
self.target_repo.resume_write_group(resume_tokens)
4270
self.target_repo.start_write_group()
4273
# locked_insert_stream performs a commit|suspend.
4274
return self._locked_insert_stream(stream, src_format,
4277
self.target_repo.abort_write_group(suppress_errors=True)
4280
self.target_repo.unlock()
4282
def _locked_insert_stream(self, stream, src_format, is_resume):
4283
to_serializer = self.target_repo._format._serializer
4284
src_serializer = src_format._serializer
4286
if to_serializer == src_serializer:
4287
# If serializers match and the target is a pack repository, set the
4288
# write cache size on the new pack. This avoids poor performance
4289
# on transports where append is unbuffered (such as
4290
# RemoteTransport). This is safe to do because nothing should read
4291
# back from the target repository while a stream with matching
4292
# serialization is being inserted.
4293
# The exception is that a delta record from the source that should
4294
# be a fulltext may need to be expanded by the target (see
4295
# test_fetch_revisions_with_deltas_into_pack); but we take care to
4296
# explicitly flush any buffered writes first in that rare case.
4298
new_pack = self.target_repo._pack_collection._new_pack
4299
except AttributeError:
4300
# Not a pack repository
4303
new_pack.set_write_cache_size(1024*1024)
4304
for substream_type, substream in stream:
4305
if 'stream' in debug.debug_flags:
4306
mutter('inserting substream: %s', substream_type)
4307
if substream_type == 'texts':
4308
self.target_repo.texts.insert_record_stream(substream)
4309
elif substream_type == 'inventories':
4310
if src_serializer == to_serializer:
4311
self.target_repo.inventories.insert_record_stream(
4314
self._extract_and_insert_inventories(
4315
substream, src_serializer)
4316
elif substream_type == 'inventory-deltas':
4317
self._extract_and_insert_inventory_deltas(
4318
substream, src_serializer)
4319
elif substream_type == 'chk_bytes':
4320
# XXX: This doesn't support conversions, as it assumes the
4321
# conversion was done in the fetch code.
4322
self.target_repo.chk_bytes.insert_record_stream(substream)
4323
elif substream_type == 'revisions':
4324
# This may fallback to extract-and-insert more often than
4325
# required if the serializers are different only in terms of
4327
if src_serializer == to_serializer:
4328
self.target_repo.revisions.insert_record_stream(substream)
4330
self._extract_and_insert_revisions(substream,
4332
elif substream_type == 'signatures':
4333
self.target_repo.signatures.insert_record_stream(substream)
4335
raise AssertionError('kaboom! %s' % (substream_type,))
4336
# Done inserting data, and the missing_keys calculations will try to
4337
# read back from the inserted data, so flush the writes to the new pack
4338
# (if this is pack format).
4339
if new_pack is not None:
4340
new_pack._write_data('', flush=True)
4341
# Find all the new revisions (including ones from resume_tokens)
4342
missing_keys = self.target_repo.get_missing_parent_inventories(
4343
check_for_missing_texts=is_resume)
4345
for prefix, versioned_file in (
4346
('texts', self.target_repo.texts),
4347
('inventories', self.target_repo.inventories),
4348
('revisions', self.target_repo.revisions),
4349
('signatures', self.target_repo.signatures),
4350
('chk_bytes', self.target_repo.chk_bytes),
4352
if versioned_file is None:
4354
# TODO: key is often going to be a StaticTuple object
4355
# I don't believe we can define a method by which
4356
# (prefix,) + StaticTuple will work, though we could
4357
# define a StaticTuple.sq_concat that would allow you to
4358
# pass in either a tuple or a StaticTuple as the second
4359
# object, so instead we could have:
4360
# StaticTuple(prefix) + key here...
4361
missing_keys.update((prefix,) + key for key in
4362
versioned_file.get_missing_compression_parent_keys())
4363
except NotImplementedError:
4364
# cannot even attempt suspending, and missing would have failed
4365
# during stream insertion.
4366
missing_keys = set()
4369
# suspend the write group and tell the caller what we is
4370
# missing. We know we can suspend or else we would not have
4371
# entered this code path. (All repositories that can handle
4372
# missing keys can handle suspending a write group).
4373
write_group_tokens = self.target_repo.suspend_write_group()
4374
return write_group_tokens, missing_keys
4375
hint = self.target_repo.commit_write_group()
4376
if (to_serializer != src_serializer and
4377
self.target_repo._format.pack_compresses):
4378
self.target_repo.pack(hint=hint)
4381
def _extract_and_insert_inventory_deltas(self, substream, serializer):
4382
target_rich_root = self.target_repo._format.rich_root_data
4383
target_tree_refs = self.target_repo._format.supports_tree_reference
4384
for record in substream:
4385
# Insert the delta directly
4386
inventory_delta_bytes = record.get_bytes_as('fulltext')
4387
deserialiser = inventory_delta.InventoryDeltaDeserializer()
4389
parse_result = deserialiser.parse_text_bytes(
4390
inventory_delta_bytes)
4391
except inventory_delta.IncompatibleInventoryDelta, err:
4392
trace.mutter("Incompatible delta: %s", err.msg)
4393
raise errors.IncompatibleRevision(self.target_repo._format)
4394
basis_id, new_id, rich_root, tree_refs, inv_delta = parse_result
4395
revision_id = new_id
4396
parents = [key[0] for key in record.parents]
4397
self.target_repo.add_inventory_by_delta(
4398
basis_id, inv_delta, revision_id, parents)
4400
def _extract_and_insert_inventories(self, substream, serializer,
4402
"""Generate a new inventory versionedfile in target, converting data.
4404
The inventory is retrieved from the source, (deserializing it), and
4405
stored in the target (reserializing it in a different format).
4407
target_rich_root = self.target_repo._format.rich_root_data
4408
target_tree_refs = self.target_repo._format.supports_tree_reference
4409
for record in substream:
4410
# It's not a delta, so it must be a fulltext in the source
4411
# serializer's format.
4412
bytes = record.get_bytes_as('fulltext')
4413
revision_id = record.key[0]
4414
inv = serializer.read_inventory_from_string(bytes, revision_id)
4415
parents = [key[0] for key in record.parents]
4416
self.target_repo.add_inventory(revision_id, inv, parents)
4417
# No need to keep holding this full inv in memory when the rest of
4418
# the substream is likely to be all deltas.
4421
def _extract_and_insert_revisions(self, substream, serializer):
4422
for record in substream:
4423
bytes = record.get_bytes_as('fulltext')
4424
revision_id = record.key[0]
4425
rev = serializer.read_revision_from_string(bytes)
4426
if rev.revision_id != revision_id:
4427
raise AssertionError('wtf: %s != %s' % (rev, revision_id))
4428
self.target_repo.add_revision(revision_id, rev)
4431
if self.target_repo._format._fetch_reconcile:
4432
self.target_repo.reconcile()
4435
class StreamSource(object):
4436
"""A source of a stream for fetching between repositories."""
4438
def __init__(self, from_repository, to_format):
4439
"""Create a StreamSource streaming from from_repository."""
4440
self.from_repository = from_repository
4441
self.to_format = to_format
4442
self._record_counter = RecordCounter()
4444
def delta_on_metadata(self):
4445
"""Return True if delta's are permitted on metadata streams.
4447
That is on revisions and signatures.
4449
src_serializer = self.from_repository._format._serializer
4450
target_serializer = self.to_format._serializer
4451
return (self.to_format._fetch_uses_deltas and
4452
src_serializer == target_serializer)
4454
def _fetch_revision_texts(self, revs):
4455
# fetch signatures first and then the revision texts
4456
# may need to be a InterRevisionStore call here.
4457
from_sf = self.from_repository.signatures
4458
# A missing signature is just skipped.
4459
keys = [(rev_id,) for rev_id in revs]
4460
signatures = versionedfile.filter_absent(from_sf.get_record_stream(
4462
self.to_format._fetch_order,
4463
not self.to_format._fetch_uses_deltas))
4464
# If a revision has a delta, this is actually expanded inside the
4465
# insert_record_stream code now, which is an alternate fix for
4467
from_rf = self.from_repository.revisions
4468
revisions = from_rf.get_record_stream(
4470
self.to_format._fetch_order,
4471
not self.delta_on_metadata())
4472
return [('signatures', signatures), ('revisions', revisions)]
4474
def _generate_root_texts(self, revs):
4475
"""This will be called by get_stream between fetching weave texts and
4476
fetching the inventory weave.
4478
if self._rich_root_upgrade():
4479
return _mod_fetch.Inter1and2Helper(
4480
self.from_repository).generate_root_texts(revs)
4484
def get_stream(self, search):
4486
revs = search.get_keys()
4487
graph = self.from_repository.get_graph()
4488
revs = tsort.topo_sort(graph.get_parent_map(revs))
4489
data_to_fetch = self.from_repository.item_keys_introduced_by(revs)
4491
for knit_kind, file_id, revisions in data_to_fetch:
4492
if knit_kind != phase:
4494
# Make a new progress bar for this phase
4495
if knit_kind == "file":
4496
# Accumulate file texts
4497
text_keys.extend([(file_id, revision) for revision in
4499
elif knit_kind == "inventory":
4500
# Now copy the file texts.
4501
from_texts = self.from_repository.texts
4502
yield ('texts', from_texts.get_record_stream(
4503
text_keys, self.to_format._fetch_order,
4504
not self.to_format._fetch_uses_deltas))
4505
# Cause an error if a text occurs after we have done the
4508
# Before we process the inventory we generate the root
4509
# texts (if necessary) so that the inventories references
4511
for _ in self._generate_root_texts(revs):
4513
# we fetch only the referenced inventories because we do not
4514
# know for unselected inventories whether all their required
4515
# texts are present in the other repository - it could be
4517
for info in self._get_inventory_stream(revs):
4519
elif knit_kind == "signatures":
4520
# Nothing to do here; this will be taken care of when
4521
# _fetch_revision_texts happens.
4523
elif knit_kind == "revisions":
4524
for record in self._fetch_revision_texts(revs):
4527
raise AssertionError("Unknown knit kind %r" % knit_kind)
4529
def get_stream_for_missing_keys(self, missing_keys):
4530
# missing keys can only occur when we are byte copying and not
4531
# translating (because translation means we don't send
4532
# unreconstructable deltas ever).
4534
keys['texts'] = set()
4535
keys['revisions'] = set()
4536
keys['inventories'] = set()
4537
keys['chk_bytes'] = set()
4538
keys['signatures'] = set()
4539
for key in missing_keys:
4540
keys[key[0]].add(key[1:])
4541
if len(keys['revisions']):
4542
# If we allowed copying revisions at this point, we could end up
4543
# copying a revision without copying its required texts: a
4544
# violation of the requirements for repository integrity.
4545
raise AssertionError(
4546
'cannot copy revisions to fill in missing deltas %s' % (
4547
keys['revisions'],))
4548
for substream_kind, keys in keys.iteritems():
4549
vf = getattr(self.from_repository, substream_kind)
4550
if vf is None and keys:
4551
raise AssertionError(
4552
"cannot fill in keys for a versioned file we don't"
4553
" have: %s needs %s" % (substream_kind, keys))
4555
# No need to stream something we don't have
4557
if substream_kind == 'inventories':
4558
# Some missing keys are genuinely ghosts, filter those out.
4559
present = self.from_repository.inventories.get_parent_map(keys)
4560
revs = [key[0] for key in present]
4561
# Get the inventory stream more-or-less as we do for the
4562
# original stream; there's no reason to assume that records
4563
# direct from the source will be suitable for the sink. (Think
4564
# e.g. 2a -> 1.9-rich-root).
4565
for info in self._get_inventory_stream(revs, missing=True):
4569
# Ask for full texts always so that we don't need more round trips
4570
# after this stream.
4571
# Some of the missing keys are genuinely ghosts, so filter absent
4572
# records. The Sink is responsible for doing another check to
4573
# ensure that ghosts don't introduce missing data for future
4575
stream = versionedfile.filter_absent(vf.get_record_stream(keys,
4576
self.to_format._fetch_order, True))
4577
yield substream_kind, stream
4579
def inventory_fetch_order(self):
4580
if self._rich_root_upgrade():
4581
return 'topological'
4583
return self.to_format._fetch_order
4585
def _rich_root_upgrade(self):
4586
return (not self.from_repository._format.rich_root_data and
4587
self.to_format.rich_root_data)
4589
def _get_inventory_stream(self, revision_ids, missing=False):
4590
from_format = self.from_repository._format
4591
if (from_format.supports_chks and self.to_format.supports_chks and
4592
from_format.network_name() == self.to_format.network_name()):
4593
raise AssertionError(
4594
"this case should be handled by GroupCHKStreamSource")
4595
elif 'forceinvdeltas' in debug.debug_flags:
4596
return self._get_convertable_inventory_stream(revision_ids,
4597
delta_versus_null=missing)
4598
elif from_format.network_name() == self.to_format.network_name():
4600
return self._get_simple_inventory_stream(revision_ids,
4602
elif (not from_format.supports_chks and not self.to_format.supports_chks
4603
and from_format._serializer == self.to_format._serializer):
4604
# Essentially the same format.
4605
return self._get_simple_inventory_stream(revision_ids,
4608
# Any time we switch serializations, we want to use an
4609
# inventory-delta based approach.
4610
return self._get_convertable_inventory_stream(revision_ids,
4611
delta_versus_null=missing)
4613
def _get_simple_inventory_stream(self, revision_ids, missing=False):
4614
# NB: This currently reopens the inventory weave in source;
4615
# using a single stream interface instead would avoid this.
4616
from_weave = self.from_repository.inventories
4618
delta_closure = True
4620
delta_closure = not self.delta_on_metadata()
4621
yield ('inventories', from_weave.get_record_stream(
4622
[(rev_id,) for rev_id in revision_ids],
4623
self.inventory_fetch_order(), delta_closure))
4625
def _get_convertable_inventory_stream(self, revision_ids,
4626
delta_versus_null=False):
4627
# The two formats are sufficiently different that there is no fast
4628
# path, so we need to send just inventorydeltas, which any
4629
# sufficiently modern client can insert into any repository.
4630
# The StreamSink code expects to be able to
4631
# convert on the target, so we need to put bytes-on-the-wire that can
4632
# be converted. That means inventory deltas (if the remote is <1.19,
4633
# RemoteStreamSink will fallback to VFS to insert the deltas).
4634
yield ('inventory-deltas',
4635
self._stream_invs_as_deltas(revision_ids,
4636
delta_versus_null=delta_versus_null))
4638
def _stream_invs_as_deltas(self, revision_ids, delta_versus_null=False):
4639
"""Return a stream of inventory-deltas for the given rev ids.
4641
:param revision_ids: The list of inventories to transmit
4642
:param delta_versus_null: Don't try to find a minimal delta for this
4643
entry, instead compute the delta versus the NULL_REVISION. This
4644
effectively streams a complete inventory. Used for stuff like
4645
filling in missing parents, etc.
4647
from_repo = self.from_repository
4648
revision_keys = [(rev_id,) for rev_id in revision_ids]
4649
parent_map = from_repo.inventories.get_parent_map(revision_keys)
4650
# XXX: possibly repos could implement a more efficient iter_inv_deltas
4652
inventories = self.from_repository.iter_inventories(
4653
revision_ids, 'topological')
4654
format = from_repo._format
4655
invs_sent_so_far = set([_mod_revision.NULL_REVISION])
4656
inventory_cache = lru_cache.LRUCache(50)
4657
null_inventory = from_repo.revision_tree(
4658
_mod_revision.NULL_REVISION).inventory
4659
# XXX: ideally the rich-root/tree-refs flags would be per-revision, not
4660
# per-repo (e.g. streaming a non-rich-root revision out of a rich-root
4661
# repo back into a non-rich-root repo ought to be allowed)
4662
serializer = inventory_delta.InventoryDeltaSerializer(
4663
versioned_root=format.rich_root_data,
4664
tree_references=format.supports_tree_reference)
4665
for inv in inventories:
4666
key = (inv.revision_id,)
4667
parent_keys = parent_map.get(key, ())
4669
if not delta_versus_null and parent_keys:
4670
# The caller did not ask for complete inventories and we have
4671
# some parents that we can delta against. Make a delta against
4672
# each parent so that we can find the smallest.
4673
parent_ids = [parent_key[0] for parent_key in parent_keys]
4674
for parent_id in parent_ids:
4675
if parent_id not in invs_sent_so_far:
4676
# We don't know that the remote side has this basis, so
4679
if parent_id == _mod_revision.NULL_REVISION:
4680
parent_inv = null_inventory
4682
parent_inv = inventory_cache.get(parent_id, None)
4683
if parent_inv is None:
4684
parent_inv = from_repo.get_inventory(parent_id)
4685
candidate_delta = inv._make_delta(parent_inv)
4686
if (delta is None or
4687
len(delta) > len(candidate_delta)):
4688
delta = candidate_delta
4689
basis_id = parent_id
4691
# Either none of the parents ended up being suitable, or we
4692
# were asked to delta against NULL
4693
basis_id = _mod_revision.NULL_REVISION
4694
delta = inv._make_delta(null_inventory)
4695
invs_sent_so_far.add(inv.revision_id)
4696
inventory_cache[inv.revision_id] = inv
4697
delta_serialized = ''.join(
4698
serializer.delta_to_lines(basis_id, key[-1], delta))
4699
yield versionedfile.FulltextContentFactory(
4700
key, parent_keys, None, delta_serialized)
4703
1771
def _iter_for_revno(repo, partial_history_cache, stop_index=None,
4704
1772
stop_revision=None):
4705
1773
"""Extend the partial history to include a given index
added
removed
bzrlib/repository.py
