114
123
note(format, *args)
116
125
def snapshot_change(self, change, path):
117
if path == '' and change in ('added', 'modified'):
126
if change == 'unchanged':
128
if change == 'added' and path == '':
119
130
self._note("%s %s", change, path)
121
132
def started(self, revno, rev_id, location=None):
122
133
if location is not None:
123
location = ' to: ' + unescape_for_display(location, 'utf-8')
134
location = ' to "' + unescape_for_display(location, 'utf-8') + '"'
125
# When started was added, location was only made optional by
126
# accident. Matt Nordhoff 20071129
127
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
128
"to started.", DeprecationWarning,
131
self._note('Committing%s', location)
137
self._note('Committing revision %d%s.', revno, location)
133
139
def completed(self, revno, rev_id):
134
140
self._note('Committed revision %d.', revno)
135
# self._note goes to the console too; so while we want to log the
136
# rev_id, we can't trivially only log it. (See bug 526425). Long
137
# term we should rearrange the reporting structure, but for now
138
# we just mutter seperately. We mutter the revid and revno together
139
# so that concurrent bzr invocations won't lead to confusion.
140
mutter('Committed revid %s as revno %d.', rev_id, revno)
142
def deleted(self, path):
143
self._note('deleted %s', path)
142
def deleted(self, file_id):
143
self._note('deleted %s', file_id)
145
def escaped(self, escape_count, message):
146
self._note("replaced %d control characters in message", escape_count)
145
148
def missing(self, path):
146
149
self._note('missing %s', path)
265
235
if message_callback is None:
266
236
if message is not None:
267
237
if isinstance(message, str):
268
message = message.decode(get_user_encoding())
238
message = message.decode(bzrlib.user_encoding)
269
239
message_callback = lambda x: message
271
241
raise BzrError("The message or message_callback keyword"
272
242
" parameter is required for commit().")
274
244
self.bound_branch = None
245
self.any_entries_changed = False
275
246
self.any_entries_deleted = False
276
if exclude is not None:
277
self.exclude = sorted(
278
minimum_path_selection(exclude))
281
247
self.local = local
282
248
self.master_branch = None
249
self.master_locked = False
283
250
self.recursive = recursive
284
251
self.rev_id = None
285
# self.specific_files is None to indicate no filter, or any iterable to
286
# indicate a filter - [] means no files at all, as per iter_changes.
287
252
if specific_files is not None:
288
253
self.specific_files = sorted(
289
254
minimum_path_selection(specific_files))
291
256
self.specific_files = None
257
self.specific_file_ids = None
293
258
self.allow_pointless = allow_pointless
259
self.revprops = revprops
294
260
self.message_callback = message_callback
295
261
self.timestamp = timestamp
296
262
self.timezone = timezone
297
263
self.committer = committer
298
264
self.strict = strict
299
265
self.verbose = verbose
266
# accumulates an inventory delta to the basis entry, so we can make
267
# just the necessary updates to the workingtree's cached basis.
268
self.basis_delta = []
301
270
self.work_tree.lock_write()
302
operation.add_cleanup(self.work_tree.unlock)
303
self.parents = self.work_tree.get_parent_ids()
304
# We can use record_iter_changes IFF iter_changes is compatible with
305
# the command line parameters, and the repository has fast delta
306
# generation. See bug 347649.
307
self.use_record_iter_changes = (
309
not self.branch.repository._format.supports_tree_reference and
310
(self.branch.repository._format.fast_deltas or
311
len(self.parents) < 2))
312
271
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
313
operation.add_cleanup(self.pb.finished)
314
self.basis_revid = self.work_tree.last_revision()
315
272
self.basis_tree = self.work_tree.basis_tree()
316
273
self.basis_tree.lock_read()
317
operation.add_cleanup(self.basis_tree.unlock)
318
# Cannot commit with conflicts present.
319
if len(self.work_tree.conflicts()) > 0:
320
raise ConflictsInTree
322
# Setup the bound branch variables as needed.
323
self._check_bound_branch(operation, possible_master_transports)
325
# Check that the working tree is up to date
326
old_revno, new_revno = self._check_out_of_date_tree()
328
# Complete configuration setup
329
if reporter is not None:
330
self.reporter = reporter
331
elif self.reporter is None:
332
self.reporter = self._select_reporter()
333
if self.config is None:
334
self.config = self.branch.get_config()
336
self._set_specific_file_ids()
338
# Setup the progress bar. As the number of files that need to be
339
# committed in unknown, progress is reported as stages.
340
# We keep track of entries separately though and include that
341
# information in the progress bar during the relevant stages.
342
self.pb_stage_name = ""
343
self.pb_stage_count = 0
344
self.pb_stage_total = 5
345
if self.bound_branch:
346
self.pb_stage_total += 1
347
self.pb.show_pct = False
348
self.pb.show_spinner = False
349
self.pb.show_eta = False
350
self.pb.show_count = True
351
self.pb.show_bar = True
353
self._gather_parents()
354
# After a merge, a selected file commit is not supported.
355
# See 'bzr help merge' for an explanation as to why.
356
if len(self.parents) > 1 and self.specific_files is not None:
357
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
358
# Excludes are a form of selected file commit.
359
if len(self.parents) > 1 and self.exclude:
360
raise errors.CannotCommitSelectedFileMerge(self.exclude)
362
# Collect the changes
363
self._set_progress_stage("Collecting changes", counter=True)
364
self.builder = self.branch.get_commit_builder(self.parents,
365
self.config, timestamp, timezone, committer, self.revprops, rev_id)
368
self.builder.will_record_deletes()
369
# find the location being committed to
370
if self.bound_branch:
371
master_location = self.master_branch.base
373
master_location = self.branch.base
375
# report the start of the commit
376
self.reporter.started(new_revno, self.rev_id, master_location)
378
self._update_builder_with_changes()
379
self._check_pointless()
381
# TODO: Now the new inventory is known, check for conflicts.
382
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
383
# weave lines, because nothing should be recorded until it is known
384
# that commit will succeed.
385
self._set_progress_stage("Saving data locally")
386
self.builder.finish_inventory()
388
# Prompt the user for a commit message if none provided
389
message = message_callback(self)
390
self.message = message
392
# Add revision data to the local branch
393
self.rev_id = self.builder.commit(self.message)
396
mutter("aborting commit write group because of exception:")
397
trace.log_exception_quietly()
398
note("aborting commit write group: %r" % (e,))
402
self._process_pre_hooks(old_revno, new_revno)
404
# Upload revision data to the master.
405
# this will propagate merged revisions too if needed.
406
if self.bound_branch:
407
self._set_progress_stage("Uploading data to master branch")
408
# 'commit' to the master first so a timeout here causes the
409
# local branch to be out of date
410
self.master_branch.import_last_revision_info(
411
self.branch.repository, new_revno, self.rev_id)
413
# and now do the commit locally.
414
self.branch.set_last_revision_info(new_revno, self.rev_id)
416
# Make the working tree be up to date with the branch. This
417
# includes automatic changes scheduled to be made to the tree, such
418
# as updating its basis and unversioning paths that were missing.
419
self.work_tree.unversion(self.deleted_ids)
420
self._set_progress_stage("Updating the working tree")
421
self.work_tree.update_basis_by_delta(self.rev_id,
422
self.builder.get_basis_delta())
423
self.reporter.completed(new_revno, self.rev_id)
424
self._process_post_hooks(old_revno, new_revno)
275
# Cannot commit with conflicts present.
276
if len(self.work_tree.conflicts()) > 0:
277
raise ConflictsInTree
279
# Setup the bound branch variables as needed.
280
self._check_bound_branch()
282
# Check that the working tree is up to date
283
old_revno, new_revno = self._check_out_of_date_tree()
285
# Complete configuration setup
286
if reporter is not None:
287
self.reporter = reporter
288
elif self.reporter is None:
289
self.reporter = self._select_reporter()
290
if self.config is None:
291
self.config = self.branch.get_config()
293
# If provided, ensure the specified files are versioned
294
if self.specific_files is not None:
295
# Note: This routine is being called because it raises
296
# PathNotVersionedError as a side effect of finding the IDs. We
297
# later use the ids we found as input to the working tree
298
# inventory iterator, so we only consider those ids rather than
299
# examining the whole tree again.
300
# XXX: Dont we have filter_unversioned to do this more
302
self.specific_file_ids = tree.find_ids_across_trees(
303
specific_files, [self.basis_tree, self.work_tree])
305
# Setup the progress bar. As the number of files that need to be
306
# committed in unknown, progress is reported as stages.
307
# We keep track of entries separately though and include that
308
# information in the progress bar during the relevant stages.
309
self.pb_stage_name = ""
310
self.pb_stage_count = 0
311
self.pb_stage_total = 5
312
if self.bound_branch:
313
self.pb_stage_total += 1
314
self.pb.show_pct = False
315
self.pb.show_spinner = False
316
self.pb.show_eta = False
317
self.pb.show_count = True
318
self.pb.show_bar = True
320
# After a merge, a selected file commit is not supported.
321
# See 'bzr help merge' for an explanation as to why.
322
self.basis_inv = self.basis_tree.inventory
323
self._gather_parents()
324
if len(self.parents) > 1 and self.specific_files:
325
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
327
# Collect the changes
328
self._set_progress_stage("Collecting changes",
329
entries_title="Directory")
330
self.builder = self.branch.get_commit_builder(self.parents,
331
self.config, timestamp, timezone, committer, revprops, rev_id)
334
# find the location being committed to
335
if self.bound_branch:
336
master_location = self.master_branch.base
338
master_location = self.branch.base
340
# report the start of the commit
341
self.reporter.started(new_revno, self.rev_id, master_location)
343
self._update_builder_with_changes()
344
self._report_and_accumulate_deletes()
345
self._check_pointless()
347
# TODO: Now the new inventory is known, check for conflicts.
348
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
349
# weave lines, because nothing should be recorded until it is known
350
# that commit will succeed.
351
self._set_progress_stage("Saving data locally")
352
self.builder.finish_inventory()
354
# Prompt the user for a commit message if none provided
355
message = message_callback(self)
356
assert isinstance(message, unicode), type(message)
357
self.message = message
358
self._escape_commit_message()
360
# Add revision data to the local branch
361
self.rev_id = self.builder.commit(self.message)
367
self._process_pre_hooks(old_revno, new_revno)
369
# Upload revision data to the master.
370
# this will propagate merged revisions too if needed.
371
if self.bound_branch:
372
self._set_progress_stage("Uploading data to master branch")
373
self.master_branch.repository.fetch(self.branch.repository,
374
revision_id=self.rev_id)
375
# now the master has the revision data
376
# 'commit' to the master first so a timeout here causes the
377
# local branch to be out of date
378
self.master_branch.set_last_revision_info(new_revno,
381
# and now do the commit locally.
382
self.branch.set_last_revision_info(new_revno, self.rev_id)
384
# Make the working tree up to date with the branch
385
self._set_progress_stage("Updating the working tree")
386
rev_tree = self.builder.revision_tree()
387
# XXX: This will need to be changed if we support doing a
388
# selective commit while a merge is still pending - then we'd
389
# still have multiple parents after the commit.
390
self.work_tree.update_basis_by_delta(self.rev_id,
392
self.reporter.completed(new_revno, self.rev_id)
393
self._process_post_hooks(old_revno, new_revno)
425
396
return self.rev_id
427
398
def _select_reporter(self):
581
555
old_revno, old_revid, new_revno, self.rev_id,
582
556
tree_delta, future_tree)
559
"""Cleanup any open locks, progress bars etc."""
560
cleanups = [self._cleanup_bound_branch,
561
self.basis_tree.unlock,
562
self.work_tree.unlock,
564
found_exception = None
565
for cleanup in cleanups:
568
# we want every cleanup to run no matter what.
569
# so we have a catchall here, but we will raise the
570
# last encountered exception up the stack: and
571
# typically this will be useful enough.
574
if found_exception is not None:
575
# don't do a plan raise, because the last exception may have been
576
# trashed, e is our sure-to-work exception even though it loses the
577
# full traceback. XXX: RBC 20060421 perhaps we could check the
578
# exc_info and if its the same one do a plain raise otherwise
579
# 'raise e' as we do now.
582
def _cleanup_bound_branch(self):
583
"""Executed at the end of a try/finally to cleanup a bound branch.
585
If the branch wasn't bound, this is a no-op.
586
If it was, it resents self.branch to the local branch, instead
589
if not self.bound_branch:
591
if self.master_locked:
592
self.master_branch.unlock()
594
def _escape_commit_message(self):
595
"""Replace xml-incompatible control characters."""
596
# FIXME: RBC 20060419 this should be done by the revision
597
# serialiser not by commit. Then we can also add an unescaper
598
# in the deserializer and start roundtripping revision messages
599
# precisely. See repository_implementations/test_repository.py
601
# Python strings can include characters that can't be
602
# represented in well-formed XML; escape characters that
603
# aren't listed in the XML specification
604
# (http://www.w3.org/TR/REC-xml/#NT-Char).
605
self.message, escape_count = re.subn(
606
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
607
lambda match: match.group(0).encode('unicode_escape'),
610
self.reporter.escaped(escape_count, self.message)
584
612
def _gather_parents(self):
585
613
"""Record the parents of a merge for merge detection."""
586
# TODO: Make sure that this list doesn't contain duplicate
614
# TODO: Make sure that this list doesn't contain duplicate
587
615
# entries and the order is preserved when doing this.
588
if self.use_record_iter_changes:
590
self.basis_inv = self.basis_tree.inventory
616
self.parents = self.work_tree.get_parent_ids()
591
617
self.parent_invs = [self.basis_inv]
592
618
for revision in self.parents[1:]:
593
619
if self.branch.repository.has_revision(revision):
600
626
def _update_builder_with_changes(self):
601
627
"""Update the commit builder with the data about what has changed.
603
exclude = self.exclude
629
# Build the revision inventory.
631
# This starts by creating a new empty inventory. Depending on
632
# which files are selected for commit, and what is present in the
633
# current tree, the new inventory is populated. inventory entries
634
# which are candidates for modification have their revision set to
635
# None; inventory entries that are carried over untouched have their
636
# revision set to their prior value.
638
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
639
# results to create a new inventory at the same time, which results
640
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
604
643
specific_files = self.specific_files
605
644
mutter("Selecting files for commit with filter %s", specific_files)
608
if self.use_record_iter_changes:
609
iter_changes = self.work_tree.iter_changes(self.basis_tree,
610
specific_files=specific_files)
611
iter_changes = self._filter_iter_changes(iter_changes)
612
for file_id, path, fs_hash in self.builder.record_iter_changes(
613
self.work_tree, self.basis_revid, iter_changes):
614
self.work_tree._observed_sha1(file_id, path, fs_hash)
616
# Build the new inventory
617
self._populate_from_inventory()
618
self._record_unselected()
619
self._report_and_accumulate_deletes()
621
def _filter_iter_changes(self, iter_changes):
622
"""Process iter_changes.
624
This method reports on the changes in iter_changes to the user, and
625
converts 'missing' entries in the iter_changes iterator to 'deleted'
626
entries. 'missing' entries have their
628
:param iter_changes: An iter_changes to process.
629
:return: A generator of changes.
631
reporter = self.reporter
632
report_changes = reporter.is_verbose()
634
for change in iter_changes:
636
old_path = change[1][0]
637
new_path = change[1][1]
638
versioned = change[3][1]
640
versioned = change[3][1]
641
if kind is None and versioned:
644
reporter.missing(new_path)
645
deleted_ids.append(change[0])
646
# Reset the new path (None) and new versioned flag (False)
647
change = (change[0], (change[1][0], None), change[2],
648
(change[3][0], False)) + change[4:]
649
elif kind == 'tree-reference':
650
if self.recursive == 'down':
651
self._commit_nested_tree(change[0], change[1][1])
652
if change[3][0] or change[3][1]:
656
reporter.deleted(old_path)
657
elif old_path is None:
658
reporter.snapshot_change('added', new_path)
659
elif old_path != new_path:
660
reporter.renamed('renamed', old_path, new_path)
663
self.work_tree.branch.repository._format.rich_root_data):
664
# Don't report on changes to '' in non rich root
666
reporter.snapshot_change('modified', new_path)
667
self._next_progress_entry()
668
# Unversion IDs that were found to be deleted
669
self.deleted_ids = deleted_ids
671
def _record_unselected(self):
646
# Build the new inventory
647
self._populate_from_inventory(specific_files)
672
649
# If specific files are selected, then all un-selected files must be
673
650
# recorded in their previous state. For more details, see
674
651
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
675
if self.specific_files or self.exclude:
676
specific_files = self.specific_files or []
677
653
for path, old_ie in self.basis_inv.iter_entries():
678
654
if old_ie.file_id in self.builder.new_inventory:
679
655
# already added - skip.
681
if (is_inside_any(specific_files, path)
682
and not is_inside_any(self.exclude, path)):
683
# was inside the selected path, and not excluded - if not
684
# present it has been deleted so skip.
657
if is_inside_any(specific_files, path):
658
# was inside the selected path, if not present it has been
686
# From here down it was either not selected, or was excluded:
687
# We preserve the entry unaltered.
661
if old_ie.kind == 'directory':
662
self._next_progress_entry()
663
# not in final inv yet, was not in the selected files, so is an
664
# entry to be preserved unaltered.
688
665
ie = old_ie.copy()
689
666
# Note: specific file commits after a merge are currently
690
667
# prohibited. This test is for sanity/safety in case it's
691
668
# required after that changes.
692
669
if len(self.parents) > 1:
693
670
ie.revision = None
694
self.builder.record_entry_contents(ie, self.parent_invs, path,
695
self.basis_tree, None)
671
delta, version_recorded = self.builder.record_entry_contents(
672
ie, self.parent_invs, path, self.basis_tree, None)
674
self.any_entries_changed = True
675
if delta: self.basis_delta.append(delta)
697
677
def _report_and_accumulate_deletes(self):
698
if (isinstance(self.basis_inv, Inventory)
699
and isinstance(self.builder.new_inventory, Inventory)):
700
# the older Inventory classes provide a _byid dict, and building a
701
# set from the keys of this dict is substantially faster than even
702
# getting a set of ids from the inventory
704
# <lifeless> set(dict) is roughly the same speed as
705
# set(iter(dict)) and both are significantly slower than
707
deleted_ids = set(self.basis_inv._byid.keys()) - \
708
set(self.builder.new_inventory._byid.keys())
710
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
678
# XXX: Could the list of deleted paths and ids be instead taken from
679
# _populate_from_inventory?
680
deleted_ids = set(self.basis_inv._byid.keys()) - \
681
set(self.builder.new_inventory._byid.keys())
712
683
self.any_entries_deleted = True
713
deleted = [(self.basis_tree.id2path(file_id), file_id)
684
deleted = [(self.basis_inv.id2path(file_id), file_id)
714
685
for file_id in deleted_ids]
716
687
# XXX: this is not quite directory-order sorting
717
688
for path, file_id in deleted:
718
self.builder.record_delete(path, file_id)
689
self.basis_delta.append((path, None, file_id, None))
719
690
self.reporter.deleted(path)
721
def _check_strict(self):
722
# XXX: when we use iter_changes this would likely be faster if
723
# iter_changes would check for us (even in the presence of
692
def _populate_from_inventory(self, specific_files):
693
"""Populate the CommitBuilder by walking the working tree inventory."""
726
695
# raise an exception as soon as we find a single unknown.
727
696
for unknown in self.work_tree.unknowns():
728
697
raise StrictCommitFailed()
730
def _populate_from_inventory(self):
731
"""Populate the CommitBuilder by walking the working tree inventory."""
732
# Build the revision inventory.
734
# This starts by creating a new empty inventory. Depending on
735
# which files are selected for commit, and what is present in the
736
# current tree, the new inventory is populated. inventory entries
737
# which are candidates for modification have their revision set to
738
# None; inventory entries that are carried over untouched have their
739
# revision set to their prior value.
741
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
742
# results to create a new inventory at the same time, which results
743
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
746
specific_files = self.specific_files
747
exclude = self.exclude
748
699
report_changes = self.reporter.is_verbose()
750
# A tree of paths that have been deleted. E.g. if foo/bar has been
751
# deleted, then we have {'foo':{'bar':{}}}
753
# XXX: Note that entries may have the wrong kind because the entry does
754
# not reflect the status on disk.
701
deleted_paths = set()
755
702
work_inv = self.work_tree.inventory
756
# NB: entries will include entries within the excluded ids/paths
757
# because iter_entries_by_dir has no 'exclude' facility today.
703
assert work_inv.root is not None
704
# XXX: Note that entries may have the wrong kind.
758
705
entries = work_inv.iter_entries_by_dir(
759
706
specific_file_ids=self.specific_file_ids, yield_parents=True)
760
707
for path, existing_ie in entries:
762
709
name = existing_ie.name
763
710
parent_id = existing_ie.parent_id
764
711
kind = existing_ie.kind
712
if kind == 'directory':
713
self._next_progress_entry()
765
714
# Skip files that have been deleted from the working tree.
766
# The deleted path ids are also recorded so they can be explicitly
769
path_segments = splitpath(path)
770
deleted_dict = deleted_paths
771
for segment in path_segments:
772
deleted_dict = deleted_dict.get(segment, None)
774
# We either took a path not present in the dict
775
# (deleted_dict was None), or we've reached an empty
776
# child dir in the dict, so are now a sub-path.
780
if deleted_dict is not None:
781
# the path has a deleted parent, do not add it.
783
if exclude and is_inside_any(exclude, path):
784
# Skip excluded paths. Excluded paths are processed by
785
# _update_builder_with_changes.
715
# The deleted files/directories are also recorded so they
716
# can be explicitly unversioned later. Note that when a
717
# filter of specific files is given, we must only skip/record
718
# deleted files matching that filter.
719
if is_inside_any(deleted_paths, path):
787
721
content_summary = self.work_tree.path_content_summary(path)
788
kind = content_summary[0]
789
# Note that when a filter of specific files is given, we must only
790
# skip/record deleted files matching that filter.
791
722
if not specific_files or is_inside_any(specific_files, path):
792
if kind == 'missing':
793
if not deleted_paths:
794
# path won't have been split yet.
795
path_segments = splitpath(path)
796
deleted_dict = deleted_paths
797
for segment in path_segments:
798
deleted_dict = deleted_dict.setdefault(segment, {})
723
if content_summary[0] == 'missing':
724
deleted_paths.add(path)
799
725
self.reporter.missing(path)
800
self._next_progress_entry()
801
726
deleted_ids.append(file_id)
803
728
# TODO: have the builder do the nested commit just-in-time IF and
804
729
# only if needed.
805
if kind == 'tree-reference':
730
if content_summary[0] == 'tree-reference':
806
731
# enforce repository nested tree policy.
807
732
if (not self.work_tree.supports_tree_reference() or
808
733
# repository does not support it either.
809
734
not self.branch.repository._format.supports_tree_reference):
811
content_summary = (kind, None, None, None)
812
elif self.recursive == 'down':
735
content_summary = ('directory',) + content_summary[1:]
736
kind = content_summary[0]
737
# TODO: specific_files filtering before nested tree processing
738
if kind == 'tree-reference':
739
if self.recursive == 'down':
813
740
nested_revision_id = self._commit_nested_tree(
815
content_summary = (kind, None, None, nested_revision_id)
742
content_summary = content_summary[:3] + (
817
nested_revision_id = self.work_tree.get_reference_revision(file_id)
818
content_summary = (kind, None, None, nested_revision_id)
745
content_summary = content_summary[:3] + (
746
self.work_tree.get_reference_revision(file_id),)
820
748
# Record an entry for this item
821
749
# Note: I don't particularly want to have the existing_ie
added
removed
bzrlib/commit.py
