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)
302
235
if message_callback is None:
303
236
if message is not None:
304
237
if isinstance(message, str):
305
message = message.decode(get_user_encoding())
238
message = message.decode(bzrlib.user_encoding)
306
239
message_callback = lambda x: message
308
241
raise BzrError("The message or message_callback keyword"
309
242
" parameter is required for commit().")
311
244
self.bound_branch = None
245
self.any_entries_changed = False
312
246
self.any_entries_deleted = False
313
if exclude is not None:
314
self.exclude = sorted(
315
minimum_path_selection(exclude))
318
247
self.local = local
319
248
self.master_branch = None
249
self.master_locked = False
320
250
self.recursive = recursive
321
251
self.rev_id = None
322
# self.specific_files is None to indicate no filter, or any iterable to
323
# indicate a filter - [] means no files at all, as per iter_changes.
324
252
if specific_files is not None:
325
253
self.specific_files = sorted(
326
254
minimum_path_selection(specific_files))
328
256
self.specific_files = None
257
self.specific_file_ids = None
330
258
self.allow_pointless = allow_pointless
259
self.revprops = revprops
331
260
self.message_callback = message_callback
332
261
self.timestamp = timestamp
333
262
self.timezone = timezone
334
263
self.committer = committer
335
264
self.strict = strict
336
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 = []
338
270
self.work_tree.lock_write()
339
operation.add_cleanup(self.work_tree.unlock)
340
self.parents = self.work_tree.get_parent_ids()
341
# We can use record_iter_changes IFF iter_changes is compatible with
342
# the command line parameters, and the repository has fast delta
343
# generation. See bug 347649.
344
self.use_record_iter_changes = (
346
not self.branch.repository._format.supports_tree_reference and
347
(self.branch.repository._format.fast_deltas or
348
len(self.parents) < 2))
349
271
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
350
operation.add_cleanup(self.pb.finished)
351
272
self.basis_revid = self.work_tree.last_revision()
352
273
self.basis_tree = self.work_tree.basis_tree()
353
274
self.basis_tree.lock_read()
354
operation.add_cleanup(self.basis_tree.unlock)
355
# Cannot commit with conflicts present.
356
if len(self.work_tree.conflicts()) > 0:
357
raise ConflictsInTree
359
# Setup the bound branch variables as needed.
360
self._check_bound_branch(operation, possible_master_transports)
362
# Check that the working tree is up to date
363
old_revno, new_revno = self._check_out_of_date_tree()
365
# Complete configuration setup
366
if reporter is not None:
367
self.reporter = reporter
368
elif self.reporter is None:
369
self.reporter = self._select_reporter()
370
if self.config is None:
371
self.config = self.branch.get_config()
373
self._set_specific_file_ids()
375
# Setup the progress bar. As the number of files that need to be
376
# committed in unknown, progress is reported as stages.
377
# We keep track of entries separately though and include that
378
# information in the progress bar during the relevant stages.
379
self.pb_stage_name = ""
380
self.pb_stage_count = 0
381
self.pb_stage_total = 5
382
if self.bound_branch:
383
self.pb_stage_total += 1
384
self.pb.show_pct = False
385
self.pb.show_spinner = False
386
self.pb.show_eta = False
387
self.pb.show_count = True
388
self.pb.show_bar = True
390
self._gather_parents()
391
# After a merge, a selected file commit is not supported.
392
# See 'bzr help merge' for an explanation as to why.
393
if len(self.parents) > 1 and self.specific_files is not None:
394
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
395
# Excludes are a form of selected file commit.
396
if len(self.parents) > 1 and self.exclude:
397
raise errors.CannotCommitSelectedFileMerge(self.exclude)
399
# Collect the changes
400
self._set_progress_stage("Collecting changes", counter=True)
401
self.builder = self.branch.get_commit_builder(self.parents,
402
self.config, timestamp, timezone, committer, self.revprops, rev_id)
405
self.builder.will_record_deletes()
406
# find the location being committed to
407
if self.bound_branch:
408
master_location = self.master_branch.base
410
master_location = self.branch.base
412
# report the start of the commit
413
self.reporter.started(new_revno, self.rev_id, master_location)
415
self._update_builder_with_changes()
416
self._check_pointless()
418
# TODO: Now the new inventory is known, check for conflicts.
419
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
420
# weave lines, because nothing should be recorded until it is known
421
# that commit will succeed.
422
self._set_progress_stage("Saving data locally")
423
self.builder.finish_inventory()
425
# Prompt the user for a commit message if none provided
426
message = message_callback(self)
427
self.message = message
429
# Add revision data to the local branch
430
self.rev_id = self.builder.commit(self.message)
433
mutter("aborting commit write group because of exception:")
434
trace.log_exception_quietly()
435
note("aborting commit write group: %r" % (e,))
439
self._process_pre_hooks(old_revno, new_revno)
441
# Upload revision data to the master.
442
# this will propagate merged revisions too if needed.
443
if self.bound_branch:
444
self._set_progress_stage("Uploading data to master branch")
445
# 'commit' to the master first so a timeout here causes the
446
# local branch to be out of date
447
self.master_branch.import_last_revision_info(
448
self.branch.repository, new_revno, self.rev_id)
450
# and now do the commit locally.
451
self.branch.set_last_revision_info(new_revno, self.rev_id)
453
# Make the working tree be up to date with the branch. This
454
# includes automatic changes scheduled to be made to the tree, such
455
# as updating its basis and unversioning paths that were missing.
456
self.work_tree.unversion(self.deleted_ids)
457
self._set_progress_stage("Updating the working tree")
458
self.work_tree.update_basis_by_delta(self.rev_id,
459
self.builder.get_basis_delta())
460
self.reporter.completed(new_revno, self.rev_id)
461
self._process_post_hooks(old_revno, new_revno)
276
# Cannot commit with conflicts present.
277
if len(self.work_tree.conflicts()) > 0:
278
raise ConflictsInTree
280
# Setup the bound branch variables as needed.
281
self._check_bound_branch()
283
# Check that the working tree is up to date
284
old_revno, new_revno = self._check_out_of_date_tree()
286
# Complete configuration setup
287
if reporter is not None:
288
self.reporter = reporter
289
elif self.reporter is None:
290
self.reporter = self._select_reporter()
291
if self.config is None:
292
self.config = self.branch.get_config()
294
# If provided, ensure the specified files are versioned
295
if self.specific_files is not None:
296
# Note: This routine is being called because it raises
297
# PathNotVersionedError as a side effect of finding the IDs. We
298
# later use the ids we found as input to the working tree
299
# inventory iterator, so we only consider those ids rather than
300
# examining the whole tree again.
301
# XXX: Dont we have filter_unversioned to do this more
303
self.specific_file_ids = tree.find_ids_across_trees(
304
specific_files, [self.basis_tree, self.work_tree])
306
# Setup the progress bar. As the number of files that need to be
307
# committed in unknown, progress is reported as stages.
308
# We keep track of entries separately though and include that
309
# information in the progress bar during the relevant stages.
310
self.pb_stage_name = ""
311
self.pb_stage_count = 0
312
self.pb_stage_total = 5
313
if self.bound_branch:
314
self.pb_stage_total += 1
315
self.pb.show_pct = False
316
self.pb.show_spinner = False
317
self.pb.show_eta = False
318
self.pb.show_count = True
319
self.pb.show_bar = True
321
# After a merge, a selected file commit is not supported.
322
# See 'bzr help merge' for an explanation as to why.
323
self.basis_inv = self.basis_tree.inventory
324
self._gather_parents()
325
if len(self.parents) > 1 and self.specific_files:
326
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
328
# Collect the changes
329
self._set_progress_stage("Collecting changes",
330
entries_title="Directory")
331
self.builder = self.branch.get_commit_builder(self.parents,
332
self.config, timestamp, timezone, committer, revprops, rev_id)
335
# find the location being committed to
336
if self.bound_branch:
337
master_location = self.master_branch.base
339
master_location = self.branch.base
341
# report the start of the commit
342
self.reporter.started(new_revno, self.rev_id, master_location)
344
self._update_builder_with_changes()
345
self._report_and_accumulate_deletes()
346
self._check_pointless()
348
# TODO: Now the new inventory is known, check for conflicts.
349
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
350
# weave lines, because nothing should be recorded until it is known
351
# that commit will succeed.
352
self._set_progress_stage("Saving data locally")
353
self.builder.finish_inventory()
355
# Prompt the user for a commit message if none provided
356
message = message_callback(self)
357
assert isinstance(message, unicode), type(message)
358
self.message = message
359
self._escape_commit_message()
361
# Add revision data to the local branch
362
self.rev_id = self.builder.commit(self.message)
368
self._process_pre_hooks(old_revno, new_revno)
370
# Upload revision data to the master.
371
# this will propagate merged revisions too if needed.
372
if self.bound_branch:
373
self._set_progress_stage("Uploading data to master branch")
374
self.master_branch.repository.fetch(self.branch.repository,
375
revision_id=self.rev_id)
376
# now the master has the revision data
377
# 'commit' to the master first so a timeout here causes the
378
# local branch to be out of date
379
self.master_branch.set_last_revision_info(new_revno,
382
# and now do the commit locally.
383
self.branch.set_last_revision_info(new_revno, self.rev_id)
385
# Make the working tree up to date with the branch
386
self._set_progress_stage("Updating the working tree")
387
rev_tree = self.builder.revision_tree()
388
# XXX: This will need to be changed if we support doing a
389
# selective commit while a merge is still pending - then we'd
390
# still have multiple parents after the commit.
392
# XXX: update_basis_by_delta is slower at present because it works
393
# on inventories, so this is not active until there's a native
394
# dirstate implementation.
395
## self.work_tree.update_basis_by_delta(self.rev_id,
396
## self._basis_delta)
397
self.work_tree.set_parent_trees([(self.rev_id, rev_tree)])
398
self.reporter.completed(new_revno, self.rev_id)
399
self._process_post_hooks(old_revno, new_revno)
462
402
return self.rev_id
464
404
def _select_reporter(self):
618
560
old_revno, old_revid, new_revno, self.rev_id,
619
561
tree_delta, future_tree)
564
"""Cleanup any open locks, progress bars etc."""
565
cleanups = [self._cleanup_bound_branch,
566
self.basis_tree.unlock,
567
self.work_tree.unlock,
569
found_exception = None
570
for cleanup in cleanups:
573
# we want every cleanup to run no matter what.
574
# so we have a catchall here, but we will raise the
575
# last encountered exception up the stack: and
576
# typically this will be useful enough.
579
if found_exception is not None:
580
# don't do a plan raise, because the last exception may have been
581
# trashed, e is our sure-to-work exception even though it loses the
582
# full traceback. XXX: RBC 20060421 perhaps we could check the
583
# exc_info and if its the same one do a plain raise otherwise
584
# 'raise e' as we do now.
587
def _cleanup_bound_branch(self):
588
"""Executed at the end of a try/finally to cleanup a bound branch.
590
If the branch wasn't bound, this is a no-op.
591
If it was, it resents self.branch to the local branch, instead
594
if not self.bound_branch:
596
if self.master_locked:
597
self.master_branch.unlock()
599
def _escape_commit_message(self):
600
"""Replace xml-incompatible control characters."""
601
# FIXME: RBC 20060419 this should be done by the revision
602
# serialiser not by commit. Then we can also add an unescaper
603
# in the deserializer and start roundtripping revision messages
604
# precisely. See repository_implementations/test_repository.py
606
# Python strings can include characters that can't be
607
# represented in well-formed XML; escape characters that
608
# aren't listed in the XML specification
609
# (http://www.w3.org/TR/REC-xml/#NT-Char).
610
self.message, escape_count = re.subn(
611
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
612
lambda match: match.group(0).encode('unicode_escape'),
615
self.reporter.escaped(escape_count, self.message)
621
617
def _gather_parents(self):
622
618
"""Record the parents of a merge for merge detection."""
623
# TODO: Make sure that this list doesn't contain duplicate
619
# TODO: Make sure that this list doesn't contain duplicate
624
620
# entries and the order is preserved when doing this.
625
if self.use_record_iter_changes:
627
self.basis_inv = self.basis_tree.inventory
621
self.parents = self.work_tree.get_parent_ids()
628
622
self.parent_invs = [self.basis_inv]
629
623
for revision in self.parents[1:]:
630
624
if self.branch.repository.has_revision(revision):
637
631
def _update_builder_with_changes(self):
638
632
"""Update the commit builder with the data about what has changed.
640
exclude = self.exclude
634
# Build the revision inventory.
636
# This starts by creating a new empty inventory. Depending on
637
# which files are selected for commit, and what is present in the
638
# current tree, the new inventory is populated. inventory entries
639
# which are candidates for modification have their revision set to
640
# None; inventory entries that are carried over untouched have their
641
# revision set to their prior value.
643
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
644
# results to create a new inventory at the same time, which results
645
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
641
648
specific_files = self.specific_files
642
649
mutter("Selecting files for commit with filter %s", specific_files)
645
if self.use_record_iter_changes:
646
iter_changes = self.work_tree.iter_changes(self.basis_tree,
647
specific_files=specific_files)
648
iter_changes = self._filter_iter_changes(iter_changes)
649
for file_id, path, fs_hash in self.builder.record_iter_changes(
650
self.work_tree, self.basis_revid, iter_changes):
651
self.work_tree._observed_sha1(file_id, path, fs_hash)
653
# Build the new inventory
654
self._populate_from_inventory()
655
self._record_unselected()
656
self._report_and_accumulate_deletes()
658
def _filter_iter_changes(self, iter_changes):
659
"""Process iter_changes.
661
This method reports on the changes in iter_changes to the user, and
662
converts 'missing' entries in the iter_changes iterator to 'deleted'
663
entries. 'missing' entries have their
665
:param iter_changes: An iter_changes to process.
666
:return: A generator of changes.
668
reporter = self.reporter
669
report_changes = reporter.is_verbose()
671
for change in iter_changes:
673
old_path = change[1][0]
674
new_path = change[1][1]
675
versioned = change[3][1]
677
versioned = change[3][1]
678
if kind is None and versioned:
681
reporter.missing(new_path)
682
deleted_ids.append(change[0])
683
# Reset the new path (None) and new versioned flag (False)
684
change = (change[0], (change[1][0], None), change[2],
685
(change[3][0], False)) + change[4:]
686
elif kind == 'tree-reference':
687
if self.recursive == 'down':
688
self._commit_nested_tree(change[0], change[1][1])
689
if change[3][0] or change[3][1]:
693
reporter.deleted(old_path)
694
elif old_path is None:
695
reporter.snapshot_change('added', new_path)
696
elif old_path != new_path:
697
reporter.renamed('renamed', old_path, new_path)
700
self.work_tree.branch.repository._format.rich_root_data):
701
# Don't report on changes to '' in non rich root
703
reporter.snapshot_change('modified', new_path)
704
self._next_progress_entry()
705
# Unversion IDs that were found to be deleted
706
self.deleted_ids = deleted_ids
708
def _record_unselected(self):
651
# Build the new inventory
652
self._populate_from_inventory(specific_files)
709
654
# If specific files are selected, then all un-selected files must be
710
655
# recorded in their previous state. For more details, see
711
656
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
712
if self.specific_files or self.exclude:
713
specific_files = self.specific_files or []
714
658
for path, old_ie in self.basis_inv.iter_entries():
715
659
if old_ie.file_id in self.builder.new_inventory:
716
660
# already added - skip.
718
if (is_inside_any(specific_files, path)
719
and not is_inside_any(self.exclude, path)):
720
# was inside the selected path, and not excluded - if not
721
# present it has been deleted so skip.
662
if is_inside_any(specific_files, path):
663
# was inside the selected path, if not present it has been
723
# From here down it was either not selected, or was excluded:
724
# We preserve the entry unaltered.
666
if old_ie.kind == 'directory':
667
self._next_progress_entry()
668
# not in final inv yet, was not in the selected files, so is an
669
# entry to be preserved unaltered.
725
670
ie = old_ie.copy()
726
671
# Note: specific file commits after a merge are currently
727
672
# prohibited. This test is for sanity/safety in case it's
728
673
# required after that changes.
729
674
if len(self.parents) > 1:
730
675
ie.revision = None
731
self.builder.record_entry_contents(ie, self.parent_invs, path,
732
self.basis_tree, None)
676
delta, version_recorded = self.builder.record_entry_contents(
677
ie, self.parent_invs, path, self.basis_tree, None)
679
self.any_entries_changed = True
680
if delta: self._basis_delta.append(delta)
734
682
def _report_and_accumulate_deletes(self):
735
if (isinstance(self.basis_inv, Inventory)
736
and isinstance(self.builder.new_inventory, Inventory)):
737
# the older Inventory classes provide a _byid dict, and building a
738
# set from the keys of this dict is substantially faster than even
739
# getting a set of ids from the inventory
741
# <lifeless> set(dict) is roughly the same speed as
742
# set(iter(dict)) and both are significantly slower than
744
deleted_ids = set(self.basis_inv._byid.keys()) - \
745
set(self.builder.new_inventory._byid.keys())
747
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
683
# XXX: Could the list of deleted paths and ids be instead taken from
684
# _populate_from_inventory?
685
deleted_ids = set(self.basis_inv._byid.keys()) - \
686
set(self.builder.new_inventory._byid.keys())
749
688
self.any_entries_deleted = True
750
689
deleted = [(self.basis_tree.id2path(file_id), file_id)
753
692
# XXX: this is not quite directory-order sorting
754
693
for path, file_id in deleted:
755
self.builder.record_delete(path, file_id)
694
self._basis_delta.append((path, None, file_id, None))
756
695
self.reporter.deleted(path)
758
def _check_strict(self):
759
# XXX: when we use iter_changes this would likely be faster if
760
# iter_changes would check for us (even in the presence of
697
def _populate_from_inventory(self, specific_files):
698
"""Populate the CommitBuilder by walking the working tree inventory."""
763
700
# raise an exception as soon as we find a single unknown.
764
701
for unknown in self.work_tree.unknowns():
765
702
raise StrictCommitFailed()
767
def _populate_from_inventory(self):
768
"""Populate the CommitBuilder by walking the working tree inventory."""
769
# Build the revision inventory.
771
# This starts by creating a new empty inventory. Depending on
772
# which files are selected for commit, and what is present in the
773
# current tree, the new inventory is populated. inventory entries
774
# which are candidates for modification have their revision set to
775
# None; inventory entries that are carried over untouched have their
776
# revision set to their prior value.
778
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
779
# results to create a new inventory at the same time, which results
780
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
783
specific_files = self.specific_files
784
exclude = self.exclude
785
704
report_changes = self.reporter.is_verbose()
787
# A tree of paths that have been deleted. E.g. if foo/bar has been
788
# deleted, then we have {'foo':{'bar':{}}}
706
deleted_paths = set()
790
707
# XXX: Note that entries may have the wrong kind because the entry does
791
708
# not reflect the status on disk.
792
709
work_inv = self.work_tree.inventory
793
# NB: entries will include entries within the excluded ids/paths
794
# because iter_entries_by_dir has no 'exclude' facility today.
795
710
entries = work_inv.iter_entries_by_dir(
796
711
specific_file_ids=self.specific_file_ids, yield_parents=True)
797
712
for path, existing_ie in entries:
799
714
name = existing_ie.name
800
715
parent_id = existing_ie.parent_id
801
716
kind = existing_ie.kind
717
if kind == 'directory':
718
self._next_progress_entry()
802
719
# Skip files that have been deleted from the working tree.
803
# The deleted path ids are also recorded so they can be explicitly
806
path_segments = splitpath(path)
807
deleted_dict = deleted_paths
808
for segment in path_segments:
809
deleted_dict = deleted_dict.get(segment, None)
811
# We either took a path not present in the dict
812
# (deleted_dict was None), or we've reached an empty
813
# child dir in the dict, so are now a sub-path.
817
if deleted_dict is not None:
818
# the path has a deleted parent, do not add it.
820
if exclude and is_inside_any(exclude, path):
821
# Skip excluded paths. Excluded paths are processed by
822
# _update_builder_with_changes.
720
# The deleted files/directories are also recorded so they
721
# can be explicitly unversioned later. Note that when a
722
# filter of specific files is given, we must only skip/record
723
# deleted files matching that filter.
724
if is_inside_any(deleted_paths, path):
824
726
content_summary = self.work_tree.path_content_summary(path)
825
kind = content_summary[0]
826
# Note that when a filter of specific files is given, we must only
827
# skip/record deleted files matching that filter.
828
727
if not specific_files or is_inside_any(specific_files, path):
829
if kind == 'missing':
830
if not deleted_paths:
831
# path won't have been split yet.
832
path_segments = splitpath(path)
833
deleted_dict = deleted_paths
834
for segment in path_segments:
835
deleted_dict = deleted_dict.setdefault(segment, {})
728
if content_summary[0] == 'missing':
729
deleted_paths.add(path)
836
730
self.reporter.missing(path)
837
self._next_progress_entry()
838
731
deleted_ids.append(file_id)
840
733
# TODO: have the builder do the nested commit just-in-time IF and
841
734
# only if needed.
842
if kind == 'tree-reference':
735
if content_summary[0] == 'tree-reference':
843
736
# enforce repository nested tree policy.
844
737
if (not self.work_tree.supports_tree_reference() or
845
738
# repository does not support it either.
846
739
not self.branch.repository._format.supports_tree_reference):
848
content_summary = (kind, None, None, None)
849
elif self.recursive == 'down':
740
content_summary = ('directory',) + content_summary[1:]
741
kind = content_summary[0]
742
# TODO: specific_files filtering before nested tree processing
743
if kind == 'tree-reference':
744
if self.recursive == 'down':
850
745
nested_revision_id = self._commit_nested_tree(
852
content_summary = (kind, None, None, nested_revision_id)
747
content_summary = content_summary[:3] + (
854
nested_revision_id = self.work_tree.get_reference_revision(file_id)
855
content_summary = (kind, None, None, nested_revision_id)
750
content_summary = content_summary[:3] + (
751
self.work_tree.get_reference_revision(file_id),)
857
753
# Record an entry for this item
858
754
# Note: I don't particularly want to have the existing_ie
added
removed
bzrlib/commit.py
