247
246
if message_callback is None:
248
247
if message is not None:
249
248
if isinstance(message, str):
250
message = message.decode(get_user_encoding())
249
message = message.decode(bzrlib.user_encoding)
251
250
message_callback = lambda x: message
253
252
raise BzrError("The message or message_callback keyword"
254
253
" parameter is required for commit().")
256
255
self.bound_branch = None
256
self.any_entries_changed = False
257
257
self.any_entries_deleted = False
258
if exclude is not None:
259
self.exclude = sorted(
260
minimum_path_selection(exclude))
263
258
self.local = local
264
259
self.master_branch = None
265
260
self.master_locked = False
266
261
self.recursive = recursive
267
262
self.rev_id = None
268
# self.specific_files is None to indicate no filter, or any iterable to
269
# indicate a filter - [] means no files at all, as per iter_changes.
270
263
if specific_files is not None:
271
264
self.specific_files = sorted(
272
265
minimum_path_selection(specific_files))
274
267
self.specific_files = None
268
self.specific_file_ids = None
276
269
self.allow_pointless = allow_pointless
277
270
self.revprops = revprops
278
271
self.message_callback = message_callback
332
329
self.pb.show_count = True
333
330
self.pb.show_bar = True
335
self._gather_parents()
336
332
# After a merge, a selected file commit is not supported.
337
333
# See 'bzr help merge' for an explanation as to why.
338
if len(self.parents) > 1 and self.specific_files is not None:
334
self.basis_inv = self.basis_tree.inventory
335
self._gather_parents()
336
if len(self.parents) > 1 and self.specific_files:
339
337
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
340
# Excludes are a form of selected file commit.
341
if len(self.parents) > 1 and self.exclude:
342
raise errors.CannotCommitSelectedFileMerge(self.exclude)
344
339
# Collect the changes
345
self._set_progress_stage("Collecting changes", counter=True)
340
self._set_progress_stage("Collecting changes",
341
entries_title="Directory")
346
342
self.builder = self.branch.get_commit_builder(self.parents,
347
343
self.config, timestamp, timezone, committer, revprops, rev_id)
350
self.builder.will_record_deletes()
351
346
# find the location being committed to
352
347
if self.bound_branch:
353
348
master_location = self.master_branch.base
387
382
# this will propagate merged revisions too if needed.
388
383
if self.bound_branch:
389
384
self._set_progress_stage("Uploading data to master branch")
385
self.master_branch.repository.fetch(self.branch.repository,
386
revision_id=self.rev_id)
387
# now the master has the revision data
390
388
# 'commit' to the master first so a timeout here causes the
391
389
# local branch to be out of date
392
self.master_branch.import_last_revision_info(
393
self.branch.repository, new_revno, self.rev_id)
390
self.master_branch.set_last_revision_info(new_revno,
395
393
# and now do the commit locally.
396
394
self.branch.set_last_revision_info(new_revno, self.rev_id)
398
# Make the working tree be up to date with the branch. This
399
# includes automatic changes scheduled to be made to the tree, such
400
# as updating its basis and unversioning paths that were missing.
401
self.work_tree.unversion(self.deleted_ids)
396
# Make the working tree up to date with the branch
402
397
self._set_progress_stage("Updating the working tree")
403
398
self.work_tree.update_basis_by_delta(self.rev_id,
404
self.builder.get_basis_delta())
405
400
self.reporter.completed(new_revno, self.rev_id)
406
401
self._process_post_hooks(old_revno, new_revno)
420
415
# A merge with no effect on files
421
416
if len(self.parents) > 1:
423
# TODO: we could simplify this by using self.builder.basis_delta.
418
# TODO: we could simplify this by using self._basis_delta.
425
420
# The initial commit adds a root directory, but this in itself is not
426
421
# a worthwhile commit.
427
422
if (self.basis_revid == revision.NULL_REVISION and
428
((self.builder.new_inventory is not None and
429
len(self.builder.new_inventory) == 1) or
430
len(self.builder._basis_delta) == 1)):
423
len(self.builder.new_inventory) == 1):
431
424
raise PointlessCommit()
432
if self.builder.any_changes():
425
# If length == 1, then we only have the root entry. Which means
426
# that there is no real difference (only the root could be different)
427
# unless deletes occured, in which case the length is irrelevant.
428
if (self.any_entries_deleted or
429
(len(self.builder.new_inventory) != 1 and
430
self.any_entries_changed)):
434
432
raise PointlessCommit()
436
def _check_bound_branch(self, possible_master_transports=None):
434
def _check_bound_branch(self):
437
435
"""Check to see if the local branch is bound.
439
437
If it is bound, then most of the commit will actually be
601
598
if self.master_locked:
602
599
self.master_branch.unlock()
601
def _escape_commit_message(self):
602
"""Replace xml-incompatible control characters."""
603
# FIXME: RBC 20060419 this should be done by the revision
604
# serialiser not by commit. Then we can also add an unescaper
605
# in the deserializer and start roundtripping revision messages
606
# precisely. See repository_implementations/test_repository.py
608
# Python strings can include characters that can't be
609
# represented in well-formed XML; escape characters that
610
# aren't listed in the XML specification
611
# (http://www.w3.org/TR/REC-xml/#NT-Char).
612
self.message, escape_count = re.subn(
613
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
614
lambda match: match.group(0).encode('unicode_escape'),
617
self.reporter.escaped(escape_count, self.message)
604
619
def _gather_parents(self):
605
620
"""Record the parents of a merge for merge detection."""
606
# TODO: Make sure that this list doesn't contain duplicate
621
# TODO: Make sure that this list doesn't contain duplicate
607
622
# entries and the order is preserved when doing this.
608
if self.use_record_iter_changes:
610
self.basis_inv = self.basis_tree.inventory
623
self.parents = self.work_tree.get_parent_ids()
611
624
self.parent_invs = [self.basis_inv]
612
625
for revision in self.parents[1:]:
613
626
if self.branch.repository.has_revision(revision):
620
633
def _update_builder_with_changes(self):
621
634
"""Update the commit builder with the data about what has changed.
623
exclude = self.exclude
636
# Build the revision inventory.
638
# This starts by creating a new empty inventory. Depending on
639
# which files are selected for commit, and what is present in the
640
# current tree, the new inventory is populated. inventory entries
641
# which are candidates for modification have their revision set to
642
# None; inventory entries that are carried over untouched have their
643
# revision set to their prior value.
645
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
646
# results to create a new inventory at the same time, which results
647
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
624
650
specific_files = self.specific_files
625
651
mutter("Selecting files for commit with filter %s", specific_files)
628
if self.use_record_iter_changes:
629
iter_changes = self.work_tree.iter_changes(self.basis_tree,
630
specific_files=specific_files)
631
iter_changes = self._filter_iter_changes(iter_changes)
632
for file_id, path, fs_hash in self.builder.record_iter_changes(
633
self.work_tree, self.basis_revid, iter_changes):
634
self.work_tree._observed_sha1(file_id, path, fs_hash)
636
# Build the new inventory
637
self._populate_from_inventory()
638
self._record_unselected()
639
self._report_and_accumulate_deletes()
641
def _filter_iter_changes(self, iter_changes):
642
"""Process iter_changes.
644
This method reports on the changes in iter_changes to the user, and
645
converts 'missing' entries in the iter_changes iterator to 'deleted'
646
entries. 'missing' entries have their
648
:param iter_changes: An iter_changes to process.
649
:return: A generator of changes.
651
reporter = self.reporter
652
report_changes = reporter.is_verbose()
654
for change in iter_changes:
656
old_path = change[1][0]
657
new_path = change[1][1]
658
versioned = change[3][1]
660
versioned = change[3][1]
661
if kind is None and versioned:
664
reporter.missing(new_path)
665
deleted_ids.append(change[0])
666
# Reset the new path (None) and new versioned flag (False)
667
change = (change[0], (change[1][0], None), change[2],
668
(change[3][0], False)) + change[4:]
669
elif kind == 'tree-reference':
670
if self.recursive == 'down':
671
self._commit_nested_tree(change[0], change[1][1])
672
if change[3][0] or change[3][1]:
676
reporter.deleted(old_path)
677
elif old_path is None:
678
reporter.snapshot_change('added', new_path)
679
elif old_path != new_path:
680
reporter.renamed('renamed', old_path, new_path)
683
self.work_tree.branch.repository._format.rich_root_data):
684
# Don't report on changes to '' in non rich root
686
reporter.snapshot_change('modified', new_path)
687
self._next_progress_entry()
688
# Unversion IDs that were found to be deleted
689
self.deleted_ids = deleted_ids
691
def _record_unselected(self):
653
# Build the new inventory
654
self._populate_from_inventory(specific_files)
692
656
# If specific files are selected, then all un-selected files must be
693
657
# recorded in their previous state. For more details, see
694
658
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
695
if self.specific_files or self.exclude:
696
specific_files = self.specific_files or []
697
660
for path, old_ie in self.basis_inv.iter_entries():
698
661
if old_ie.file_id in self.builder.new_inventory:
699
662
# already added - skip.
701
if (is_inside_any(specific_files, path)
702
and not is_inside_any(self.exclude, path)):
703
# was inside the selected path, and not excluded - if not
704
# present it has been deleted so skip.
664
if is_inside_any(specific_files, path):
665
# was inside the selected path, if not present it has been
706
# From here down it was either not selected, or was excluded:
707
# We preserve the entry unaltered.
668
if old_ie.kind == 'directory':
669
self._next_progress_entry()
670
# not in final inv yet, was not in the selected files, so is an
671
# entry to be preserved unaltered.
708
672
ie = old_ie.copy()
709
673
# Note: specific file commits after a merge are currently
710
674
# prohibited. This test is for sanity/safety in case it's
711
675
# required after that changes.
712
676
if len(self.parents) > 1:
713
677
ie.revision = None
714
self.builder.record_entry_contents(ie, self.parent_invs, path,
715
self.basis_tree, None)
678
delta, version_recorded = self.builder.record_entry_contents(
679
ie, self.parent_invs, path, self.basis_tree, None)
681
self.any_entries_changed = True
682
if delta: self._basis_delta.append(delta)
717
684
def _report_and_accumulate_deletes(self):
718
if (isinstance(self.basis_inv, Inventory)
719
and isinstance(self.builder.new_inventory, Inventory)):
720
# the older Inventory classes provide a _byid dict, and building a
721
# set from the keys of this dict is substantially faster than even
722
# getting a set of ids from the inventory
724
# <lifeless> set(dict) is roughly the same speed as
725
# set(iter(dict)) and both are significantly slower than
727
deleted_ids = set(self.basis_inv._byid.keys()) - \
728
set(self.builder.new_inventory._byid.keys())
730
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
685
# XXX: Could the list of deleted paths and ids be instead taken from
686
# _populate_from_inventory?
687
deleted_ids = set(self.basis_inv._byid.keys()) - \
688
set(self.builder.new_inventory._byid.keys())
732
690
self.any_entries_deleted = True
733
691
deleted = [(self.basis_tree.id2path(file_id), file_id)
736
694
# XXX: this is not quite directory-order sorting
737
695
for path, file_id in deleted:
738
self.builder.record_delete(path, file_id)
696
self._basis_delta.append((path, None, file_id, None))
739
697
self.reporter.deleted(path)
741
def _check_strict(self):
742
# XXX: when we use iter_changes this would likely be faster if
743
# iter_changes would check for us (even in the presence of
699
def _populate_from_inventory(self, specific_files):
700
"""Populate the CommitBuilder by walking the working tree inventory."""
746
702
# raise an exception as soon as we find a single unknown.
747
703
for unknown in self.work_tree.unknowns():
748
704
raise StrictCommitFailed()
750
def _populate_from_inventory(self):
751
"""Populate the CommitBuilder by walking the working tree inventory."""
752
# Build the revision inventory.
754
# This starts by creating a new empty inventory. Depending on
755
# which files are selected for commit, and what is present in the
756
# current tree, the new inventory is populated. inventory entries
757
# which are candidates for modification have their revision set to
758
# None; inventory entries that are carried over untouched have their
759
# revision set to their prior value.
761
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
762
# results to create a new inventory at the same time, which results
763
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
766
specific_files = self.specific_files
767
exclude = self.exclude
768
706
report_changes = self.reporter.is_verbose()
770
708
# A tree of paths that have been deleted. E.g. if foo/bar has been
817
750
for segment in path_segments:
818
751
deleted_dict = deleted_dict.setdefault(segment, {})
819
752
self.reporter.missing(path)
820
self._next_progress_entry()
821
753
deleted_ids.append(file_id)
823
755
# TODO: have the builder do the nested commit just-in-time IF and
824
756
# only if needed.
825
if kind == 'tree-reference':
757
if content_summary[0] == 'tree-reference':
826
758
# enforce repository nested tree policy.
827
759
if (not self.work_tree.supports_tree_reference() or
828
760
# repository does not support it either.
829
761
not self.branch.repository._format.supports_tree_reference):
831
content_summary = (kind, None, None, None)
832
elif self.recursive == 'down':
762
content_summary = ('directory',) + content_summary[1:]
763
kind = content_summary[0]
764
# TODO: specific_files filtering before nested tree processing
765
if kind == 'tree-reference':
766
if self.recursive == 'down':
833
767
nested_revision_id = self._commit_nested_tree(
835
content_summary = (kind, None, None, nested_revision_id)
769
content_summary = content_summary[:3] + (
837
nested_revision_id = self.work_tree.get_reference_revision(file_id)
838
content_summary = (kind, None, None, nested_revision_id)
772
content_summary = content_summary[:3] + (
773
self.work_tree.get_reference_revision(file_id),)
840
775
# Record an entry for this item
841
776
# Note: I don't particularly want to have the existing_ie
908
843
change = ie.describe_change(basis_ie, ie)
909
if change in (InventoryEntry.RENAMED,
844
if change in (InventoryEntry.RENAMED,
910
845
InventoryEntry.MODIFIED_AND_RENAMED):
911
846
old_path = self.basis_inv.id2path(ie.file_id)
912
847
self.reporter.renamed(change, old_path, path)
913
self._next_progress_entry()
915
if change == 'unchanged':
917
849
self.reporter.snapshot_change(change, path)
918
self._next_progress_entry()
920
def _set_progress_stage(self, name, counter=False):
851
def _set_progress_stage(self, name, entries_title=None):
921
852
"""Set the progress stage and emit an update to the progress bar."""
922
853
self.pb_stage_name = name
923
854
self.pb_stage_count += 1
855
self.pb_entries_title = entries_title
856
if entries_title is not None:
925
857
self.pb_entries_count = 0
927
self.pb_entries_count = None
858
self.pb_entries_total = '?'
928
859
self._emit_progress()
930
861
def _next_progress_entry(self):
933
864
self._emit_progress()
935
866
def _emit_progress(self):
936
if self.pb_entries_count is not None:
937
text = "%s [%d] - Stage" % (self.pb_stage_name,
938
self.pb_entries_count)
867
if self.pb_entries_title:
868
if self.pb_entries_total == '?':
869
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
870
self.pb_entries_title, self.pb_entries_count)
872
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
873
self.pb_entries_title, self.pb_entries_count,
874
str(self.pb_entries_total))
940
text = "%s - Stage" % (self.pb_stage_name, )
876
text = "%s - Stage" % (self.pb_stage_name)
941
877
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
943
def _set_specific_file_ids(self):
944
"""populate self.specific_file_ids if we will use it."""
945
if not self.use_record_iter_changes:
946
# If provided, ensure the specified files are versioned
947
if self.specific_files is not None:
948
# Note: This routine is being called because it raises
949
# PathNotVersionedError as a side effect of finding the IDs. We
950
# later use the ids we found as input to the working tree
951
# inventory iterator, so we only consider those ids rather than
952
# examining the whole tree again.
953
# XXX: Dont we have filter_unversioned to do this more
955
self.specific_file_ids = tree.find_ids_across_trees(
956
self.specific_files, [self.basis_tree, self.work_tree])
958
self.specific_file_ids = None