298
236
self.strict = strict
299
237
self.verbose = verbose
239
if reporter is None and self.reporter is None:
240
self.reporter = NullCommitReporter()
241
elif reporter is not None:
242
self.reporter = reporter
301
244
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
245
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
246
self.basis_tree = self.work_tree.basis_tree()
316
247
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)
249
# Cannot commit with conflicts present.
250
if len(self.work_tree.conflicts()) > 0:
251
raise ConflictsInTree
253
# Setup the bound branch variables as needed.
254
self._check_bound_branch()
256
# Check that the working tree is up to date
257
old_revno, new_revno = self._check_out_of_date_tree()
259
if self.config is None:
260
self.config = self.branch.get_config()
262
# If provided, ensure the specified files are versioned
263
if specific_files is not None:
264
# Note: We don't actually need the IDs here. This routine
265
# is being called because it raises PathNotVerisonedError
266
# as a side effect of finding the IDs.
267
# XXX: Dont we have filter_unversioned to do this more
269
tree.find_ids_across_trees(specific_files,
270
[self.basis_tree, self.work_tree])
272
# Setup the progress bar. As the number of files that need to be
273
# committed in unknown, progress is reported as stages.
274
# We keep track of entries separately though and include that
275
# information in the progress bar during the relevant stages.
276
self.pb_stage_name = ""
277
self.pb_stage_count = 0
278
self.pb_stage_total = 5
279
if self.bound_branch:
280
self.pb_stage_total += 1
281
self.pb.show_pct = False
282
self.pb.show_spinner = False
283
self.pb.show_eta = False
284
self.pb.show_count = True
285
self.pb.show_bar = True
287
# After a merge, a selected file commit is not supported.
288
# See 'bzr help merge' for an explanation as to why.
289
self.basis_inv = self.basis_tree.inventory
290
self._gather_parents()
291
if len(self.parents) > 1 and self.specific_files:
292
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
294
# Collect the changes
295
self._set_progress_stage("Collecting changes",
296
entries_title="Directory")
297
self.builder = self.branch.get_commit_builder(self.parents,
298
self.config, timestamp, timezone, committer, revprops, rev_id)
301
self._update_builder_with_changes()
302
self._check_pointless()
304
# TODO: Now the new inventory is known, check for conflicts.
305
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
306
# weave lines, because nothing should be recorded until it is known
307
# that commit will succeed.
308
self._set_progress_stage("Saving data locally")
309
self.builder.finish_inventory()
311
# Prompt the user for a commit message if none provided
312
message = message_callback(self)
313
assert isinstance(message, unicode), type(message)
314
self.message = message
315
self._escape_commit_message()
317
# Add revision data to the local branch
318
self.rev_id = self.builder.commit(self.message)
324
self._process_pre_hooks(old_revno, new_revno)
326
# Upload revision data to the master.
327
# this will propagate merged revisions too if needed.
328
if self.bound_branch:
329
self._set_progress_stage("Uploading data to master branch")
330
self.master_branch.repository.fetch(self.branch.repository,
331
revision_id=self.rev_id)
332
# now the master has the revision data
333
# 'commit' to the master first so a timeout here causes the
334
# local branch to be out of date
335
self.master_branch.set_last_revision_info(new_revno,
338
# and now do the commit locally.
339
self.branch.set_last_revision_info(new_revno, self.rev_id)
341
# Make the working tree up to date with the branch
342
self._set_progress_stage("Updating the working tree")
343
rev_tree = self.builder.revision_tree()
344
self.work_tree.set_parent_trees([(self.rev_id, rev_tree)])
345
self.reporter.completed(new_revno, self.rev_id)
346
self._process_post_hooks(old_revno, new_revno)
425
349
return self.rev_id
427
def _select_reporter(self):
428
"""Select the CommitReporter to use."""
430
return NullCommitReporter()
431
return ReportCommitToLog()
351
def _any_real_changes(self):
352
"""Are there real changes between new_inventory and basis?
354
For trees without rich roots, inv.root.revision changes every commit.
355
But if that is the only change, we want to treat it as though there
358
new_entries = self.builder.new_inventory.iter_entries()
359
basis_entries = self.basis_inv.iter_entries()
360
new_path, new_root_ie = new_entries.next()
361
basis_path, basis_root_ie = basis_entries.next()
363
# This is a copy of InventoryEntry.__eq__ only leaving out .revision
364
def ie_equal_no_revision(this, other):
365
return ((this.file_id == other.file_id)
366
and (this.name == other.name)
367
and (this.symlink_target == other.symlink_target)
368
and (this.text_sha1 == other.text_sha1)
369
and (this.text_size == other.text_size)
370
and (this.text_id == other.text_id)
371
and (this.parent_id == other.parent_id)
372
and (this.kind == other.kind)
373
and (this.executable == other.executable)
374
and (this.reference_revision == other.reference_revision)
376
if not ie_equal_no_revision(new_root_ie, basis_root_ie):
379
for new_ie, basis_ie in zip(new_entries, basis_entries):
380
if new_ie != basis_ie:
383
# No actual changes present
433
386
def _check_pointless(self):
434
387
if self.allow_pointless:
581
534
old_revno, old_revid, new_revno, self.rev_id,
582
535
tree_delta, future_tree)
538
"""Cleanup any open locks, progress bars etc."""
539
cleanups = [self._cleanup_bound_branch,
540
self.basis_tree.unlock,
541
self.work_tree.unlock,
543
found_exception = None
544
for cleanup in cleanups:
547
# we want every cleanup to run no matter what.
548
# so we have a catchall here, but we will raise the
549
# last encountered exception up the stack: and
550
# typically this will be useful enough.
553
if found_exception is not None:
554
# don't do a plan raise, because the last exception may have been
555
# trashed, e is our sure-to-work exception even though it loses the
556
# full traceback. XXX: RBC 20060421 perhaps we could check the
557
# exc_info and if its the same one do a plain raise otherwise
558
# 'raise e' as we do now.
561
def _cleanup_bound_branch(self):
562
"""Executed at the end of a try/finally to cleanup a bound branch.
564
If the branch wasn't bound, this is a no-op.
565
If it was, it resents self.branch to the local branch, instead
568
if not self.bound_branch:
570
if self.master_locked:
571
self.master_branch.unlock()
573
def _escape_commit_message(self):
574
"""Replace xml-incompatible control characters."""
575
# FIXME: RBC 20060419 this should be done by the revision
576
# serialiser not by commit. Then we can also add an unescaper
577
# in the deserializer and start roundtripping revision messages
578
# precisely. See repository_implementations/test_repository.py
580
# Python strings can include characters that can't be
581
# represented in well-formed XML; escape characters that
582
# aren't listed in the XML specification
583
# (http://www.w3.org/TR/REC-xml/#NT-Char).
584
self.message, escape_count = re.subn(
585
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
586
lambda match: match.group(0).encode('unicode_escape'),
589
self.reporter.escaped(escape_count, self.message)
584
591
def _gather_parents(self):
585
592
"""Record the parents of a merge for merge detection."""
586
# TODO: Make sure that this list doesn't contain duplicate
593
# TODO: Make sure that this list doesn't contain duplicate
587
594
# entries and the order is preserved when doing this.
588
if self.use_record_iter_changes:
590
self.basis_inv = self.basis_tree.inventory
595
self.parents = self.work_tree.get_parent_ids()
591
596
self.parent_invs = [self.basis_inv]
592
597
for revision in self.parents[1:]:
593
598
if self.branch.repository.has_revision(revision):
600
605
def _update_builder_with_changes(self):
601
606
"""Update the commit builder with the data about what has changed.
603
exclude = self.exclude
608
# Build the revision inventory.
610
# This starts by creating a new empty inventory. Depending on
611
# which files are selected for commit, and what is present in the
612
# current tree, the new inventory is populated. inventory entries
613
# which are candidates for modification have their revision set to
614
# None; inventory entries that are carried over untouched have their
615
# revision set to their prior value.
617
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
618
# results to create a new inventory at the same time, which results
619
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
604
622
specific_files = self.specific_files
605
623
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):
625
# Check and warn about old CommitBuilders
626
if not self.builder.record_root_entry:
627
symbol_versioning.warn('CommitBuilders should support recording'
628
' the root entry as of bzr 0.10.', DeprecationWarning,
630
self.builder.new_inventory.add(self.basis_inv.root.copy())
632
# Build the new inventory
633
self._populate_from_inventory(specific_files)
672
635
# If specific files are selected, then all un-selected files must be
673
636
# recorded in their previous state. For more details, see
674
637
# 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
for path, old_ie in self.basis_inv.iter_entries():
678
if old_ie.file_id in self.builder.new_inventory:
679
# 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.
686
# From here down it was either not selected, or was excluded:
687
# We preserve the entry unaltered.
689
# Note: specific file commits after a merge are currently
690
# prohibited. This test is for sanity/safety in case it's
691
# required after that changes.
692
if len(self.parents) > 1:
639
for path, new_ie in self.basis_inv.iter_entries():
640
if new_ie.file_id in self.builder.new_inventory:
642
if is_inside_any(specific_files, path):
694
646
self.builder.record_entry_contents(ie, self.parent_invs, path,
695
self.basis_tree, None)
697
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)
712
self.any_entries_deleted = True
713
deleted = [(self.basis_tree.id2path(file_id), file_id)
714
for file_id in deleted_ids]
716
# XXX: this is not quite directory-order sorting
717
for path, file_id in deleted:
718
self.builder.record_delete(path, file_id)
649
# Report what was deleted. We could skip this when no deletes are
650
# detected to gain a performance win, but it arguably serves as a
651
# 'safety check' by informing the user whenever anything disappears.
652
for path, ie in self.basis_inv.iter_entries():
653
if ie.file_id not in self.builder.new_inventory:
719
654
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
656
def _populate_from_inventory(self, specific_files):
657
"""Populate the CommitBuilder by walking the working tree inventory."""
726
659
# raise an exception as soon as we find a single unknown.
727
660
for unknown in self.work_tree.unknowns():
728
661
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
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.
664
deleted_paths = set()
755
665
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.
758
entries = work_inv.iter_entries_by_dir(
759
specific_file_ids=self.specific_file_ids, yield_parents=True)
666
assert work_inv.root is not None
667
entries = work_inv.iter_entries()
668
if not self.builder.record_root_entry:
760
670
for path, existing_ie in entries:
761
671
file_id = existing_ie.file_id
762
672
name = existing_ie.name
763
673
parent_id = existing_ie.parent_id
764
674
kind = existing_ie.kind
675
if kind == 'directory':
676
self._next_progress_entry()
765
678
# 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.
679
# The deleted files/directories are also recorded so they
680
# can be explicitly unversioned later. Note that when a
681
# filter of specific files is given, we must only skip/record
682
# deleted files matching that filter.
683
if is_inside_any(deleted_paths, path):
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.
791
685
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, {})
686
if not self.work_tree.has_filename(path):
687
deleted_paths.add(path)
799
688
self.reporter.missing(path)
800
self._next_progress_entry()
801
689
deleted_ids.append(file_id)
803
# TODO: have the builder do the nested commit just-in-time IF and
805
if kind == 'tree-reference':
806
# enforce repository nested tree policy.
807
if (not self.work_tree.supports_tree_reference() or
808
# repository does not support it either.
809
not self.branch.repository._format.supports_tree_reference):
811
content_summary = (kind, None, None, None)
812
elif self.recursive == 'down':
813
nested_revision_id = self._commit_nested_tree(
815
content_summary = (kind, None, None, nested_revision_id)
817
nested_revision_id = self.work_tree.get_reference_revision(file_id)
818
content_summary = (kind, None, None, nested_revision_id)
692
kind = self.work_tree.kind(file_id)
693
# TODO: specific_files filtering before nested tree processing
694
if kind == 'tree-reference' and self.recursive == 'down':
695
self._commit_nested_tree(file_id, path)
696
except errors.NoSuchFile:
820
699
# Record an entry for this item
821
700
# Note: I don't particularly want to have the existing_ie
822
701
# parameter but the test suite currently (28-Jun-07) breaks
823
702
# without it thanks to a unicode normalisation issue. :-(
824
definitely_changed = kind != existing_ie.kind
703
definitely_changed = kind != existing_ie.kind
825
704
self._record_entry(path, file_id, specific_files, kind, name,
826
parent_id, definitely_changed, existing_ie, report_changes,
705
parent_id, definitely_changed, existing_ie)
829
707
# Unversion IDs that were found to be deleted
830
self.deleted_ids = deleted_ids
708
self.work_tree.unversion(deleted_ids)
832
710
def _commit_nested_tree(self, file_id, path):
833
711
"Commit a nested tree."
852
730
strict=self.strict, verbose=self.verbose,
853
731
local=self.local, reporter=self.reporter)
854
732
except errors.PointlessCommit:
855
return self.work_tree.get_reference_revision(file_id)
857
735
def _record_entry(self, path, file_id, specific_files, kind, name,
858
parent_id, definitely_changed, existing_ie, report_changes,
736
parent_id, definitely_changed, existing_ie=None):
860
737
"Record the new inventory entry for a path if any."
861
738
# mutter('check %s {%s}', path, file_id)
862
# mutter('%s selected for commit', path)
863
if definitely_changed or existing_ie is None:
864
ie = make_entry(kind, name, parent_id, file_id)
739
if (not specific_files or
740
is_inside_or_parent_of_any(specific_files, path)):
741
# mutter('%s selected for commit', path)
742
if definitely_changed or existing_ie is None:
743
ie = inventory.make_entry(kind, name, parent_id, file_id)
745
ie = existing_ie.copy()
866
ie = existing_ie.copy()
868
# For carried over entries we don't care about the fs hash - the repo
869
# isn't generating a sha, so we're not saving computation time.
870
_, _, fs_hash = self.builder.record_entry_contents(
871
ie, self.parent_invs, path, self.work_tree, content_summary)
748
# mutter('%s not selected for commit', path)
749
if self.basis_inv.has_id(file_id):
750
ie = self.basis_inv[file_id].copy()
752
# this entry is new and not being committed
755
self.builder.record_entry_contents(ie, self.parent_invs,
756
path, self.work_tree)
873
757
self._report_change(ie, path)
875
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
878
760
def _report_change(self, ie, path):
added
removed
bzrlib/commit.py
