123
114
note(format, *args)
125
116
def snapshot_change(self, change, path):
126
if change == 'unchanged':
128
if change == 'added' and path == '':
117
if path == '' and change in ('added', 'modified'):
130
119
self._note("%s %s", change, path)
132
121
def started(self, revno, rev_id, location=None):
133
122
if location is not None:
134
location = ' to "' + unescape_for_display(location, 'utf-8') + '"'
123
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,
137
self._note('Committing revision %d%s.', revno, location)
131
self._note('Committing%s', location)
139
133
def completed(self, revno, rev_id):
140
134
self._note('Committed revision %d.', revno)
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)
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)
148
145
def missing(self, path):
149
146
self._note('missing %s', path)
235
265
if message_callback is None:
236
266
if message is not None:
237
267
if isinstance(message, str):
238
message = message.decode(bzrlib.user_encoding)
268
message = message.decode(get_user_encoding())
239
269
message_callback = lambda x: message
241
271
raise BzrError("The message or message_callback keyword"
242
272
" parameter is required for commit().")
244
274
self.bound_branch = None
245
self.any_entries_changed = False
246
275
self.any_entries_deleted = False
276
if exclude is not None:
277
self.exclude = sorted(
278
minimum_path_selection(exclude))
247
281
self.local = local
248
282
self.master_branch = None
249
self.master_locked = False
250
283
self.recursive = recursive
251
284
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.
252
287
if specific_files is not None:
253
288
self.specific_files = sorted(
254
289
minimum_path_selection(specific_files))
256
291
self.specific_files = None
257
self.specific_file_ids = None
258
293
self.allow_pointless = allow_pointless
259
self.revprops = revprops
260
294
self.message_callback = message_callback
261
295
self.timestamp = timestamp
262
296
self.timezone = timezone
263
297
self.committer = committer
264
298
self.strict = strict
265
299
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 = []
270
301
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))
271
312
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
313
operation.add_cleanup(self.pb.finished)
272
314
self.basis_revid = self.work_tree.last_revision()
273
315
self.basis_tree = self.work_tree.basis_tree()
274
316
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)
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)
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)
402
425
return self.rev_id
404
427
def _select_reporter(self):
560
581
old_revno, old_revid, new_revno, self.rev_id,
561
582
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)
617
584
def _gather_parents(self):
618
585
"""Record the parents of a merge for merge detection."""
619
# TODO: Make sure that this list doesn't contain duplicate
586
# TODO: Make sure that this list doesn't contain duplicate
620
587
# entries and the order is preserved when doing this.
621
self.parents = self.work_tree.get_parent_ids()
588
if self.use_record_iter_changes:
590
self.basis_inv = self.basis_tree.inventory
622
591
self.parent_invs = [self.basis_inv]
623
592
for revision in self.parents[1:]:
624
593
if self.branch.repository.has_revision(revision):
631
600
def _update_builder_with_changes(self):
632
601
"""Update the commit builder with the data about what has changed.
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?
603
exclude = self.exclude
648
604
specific_files = self.specific_files
649
605
mutter("Selecting files for commit with filter %s", specific_files)
651
# Build the new inventory
652
self._populate_from_inventory(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):
654
672
# If specific files are selected, then all un-selected files must be
655
673
# recorded in their previous state. For more details, see
656
674
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
675
if self.specific_files or self.exclude:
676
specific_files = self.specific_files or []
658
677
for path, old_ie in self.basis_inv.iter_entries():
659
678
if old_ie.file_id in self.builder.new_inventory:
660
679
# already added - skip.
662
if is_inside_any(specific_files, path):
663
# was inside the selected path, if not present it has been
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.
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.
686
# From here down it was either not selected, or was excluded:
687
# We preserve the entry unaltered.
670
688
ie = old_ie.copy()
671
689
# Note: specific file commits after a merge are currently
672
690
# prohibited. This test is for sanity/safety in case it's
673
691
# required after that changes.
674
692
if len(self.parents) > 1:
675
693
ie.revision = 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)
694
self.builder.record_entry_contents(ie, self.parent_invs, path,
695
self.basis_tree, None)
682
697
def _report_and_accumulate_deletes(self):
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())
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)
688
712
self.any_entries_deleted = True
689
713
deleted = [(self.basis_tree.id2path(file_id), file_id)
692
716
# XXX: this is not quite directory-order sorting
693
717
for path, file_id in deleted:
694
self._basis_delta.append((path, None, file_id, None))
718
self.builder.record_delete(path, file_id)
695
719
self.reporter.deleted(path)
697
def _populate_from_inventory(self, specific_files):
698
"""Populate the CommitBuilder by walking the working tree inventory."""
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
700
726
# raise an exception as soon as we find a single unknown.
701
727
for unknown in self.work_tree.unknowns():
702
728
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
704
748
report_changes = self.reporter.is_verbose()
706
deleted_paths = set()
750
# A tree of paths that have been deleted. E.g. if foo/bar has been
751
# deleted, then we have {'foo':{'bar':{}}}
707
753
# XXX: Note that entries may have the wrong kind because the entry does
708
754
# not reflect the status on disk.
709
755
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.
710
758
entries = work_inv.iter_entries_by_dir(
711
759
specific_file_ids=self.specific_file_ids, yield_parents=True)
712
760
for path, existing_ie in entries:
714
762
name = existing_ie.name
715
763
parent_id = existing_ie.parent_id
716
764
kind = existing_ie.kind
717
if kind == 'directory':
718
self._next_progress_entry()
719
765
# Skip files that have been deleted from the working tree.
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):
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.
726
787
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.
727
791
if not specific_files or is_inside_any(specific_files, path):
728
if content_summary[0] == 'missing':
729
deleted_paths.add(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, {})
730
799
self.reporter.missing(path)
800
self._next_progress_entry()
731
801
deleted_ids.append(file_id)
733
803
# TODO: have the builder do the nested commit just-in-time IF and
734
804
# only if needed.
735
if content_summary[0] == 'tree-reference':
805
if kind == 'tree-reference':
736
806
# enforce repository nested tree policy.
737
807
if (not self.work_tree.supports_tree_reference() or
738
808
# repository does not support it either.
739
809
not self.branch.repository._format.supports_tree_reference):
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':
811
content_summary = (kind, None, None, None)
812
elif self.recursive == 'down':
745
813
nested_revision_id = self._commit_nested_tree(
747
content_summary = content_summary[:3] + (
815
content_summary = (kind, None, None, nested_revision_id)
750
content_summary = content_summary[:3] + (
751
self.work_tree.get_reference_revision(file_id),)
817
nested_revision_id = self.work_tree.get_reference_revision(file_id)
818
content_summary = (kind, None, None, nested_revision_id)
753
820
# Record an entry for this item
754
821
# Note: I don't particularly want to have the existing_ie
added
removed
bzrlib/commit.py
