307
236
self.strict = strict
308
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
310
244
self.work_tree.lock_write()
311
operation.add_cleanup(self.work_tree.unlock)
312
self.parents = self.work_tree.get_parent_ids()
313
# We can use record_iter_changes IFF iter_changes is compatible with
314
# the command line parameters, and the repository has fast delta
315
# generation. See bug 347649.
316
self.use_record_iter_changes = (
318
not self.branch.repository._format.supports_tree_reference and
319
(self.branch.repository._format.fast_deltas or
320
len(self.parents) < 2))
321
245
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
322
operation.add_cleanup(self.pb.finished)
323
self.basis_revid = self.work_tree.last_revision()
324
246
self.basis_tree = self.work_tree.basis_tree()
325
247
self.basis_tree.lock_read()
326
operation.add_cleanup(self.basis_tree.unlock)
327
# Cannot commit with conflicts present.
328
if len(self.work_tree.conflicts()) > 0:
329
raise ConflictsInTree
331
# Setup the bound branch variables as needed.
332
self._check_bound_branch(operation, possible_master_transports)
334
# Check that the working tree is up to date
335
old_revno, new_revno = self._check_out_of_date_tree()
337
# Complete configuration setup
338
if reporter is not None:
339
self.reporter = reporter
340
elif self.reporter is None:
341
self.reporter = self._select_reporter()
342
if self.config is None:
343
self.config = self.branch.get_config()
345
self._set_specific_file_ids()
347
# Setup the progress bar. As the number of files that need to be
348
# committed in unknown, progress is reported as stages.
349
# We keep track of entries separately though and include that
350
# information in the progress bar during the relevant stages.
351
self.pb_stage_name = ""
352
self.pb_stage_count = 0
353
self.pb_stage_total = 5
354
if self.bound_branch:
355
self.pb_stage_total += 1
356
self.pb.show_pct = False
357
self.pb.show_spinner = False
358
self.pb.show_eta = False
359
self.pb.show_count = True
360
self.pb.show_bar = True
362
self._gather_parents()
363
# After a merge, a selected file commit is not supported.
364
# See 'bzr help merge' for an explanation as to why.
365
if len(self.parents) > 1 and self.specific_files is not None:
366
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
367
# Excludes are a form of selected file commit.
368
if len(self.parents) > 1 and self.exclude:
369
raise errors.CannotCommitSelectedFileMerge(self.exclude)
371
# Collect the changes
372
self._set_progress_stage("Collecting changes", counter=True)
373
self.builder = self.branch.get_commit_builder(self.parents,
374
self.config, timestamp, timezone, committer, revprops, rev_id)
377
self.builder.will_record_deletes()
378
# find the location being committed to
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 = 4
379
279
if self.bound_branch:
380
master_location = self.master_branch.base
382
master_location = self.branch.base
384
# report the start of the commit
385
self.reporter.started(new_revno, self.rev_id, master_location)
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)
387
299
self._update_builder_with_changes()
388
300
self._check_pointless()
397
309
# Prompt the user for a commit message if none provided
398
310
message = message_callback(self)
311
assert isinstance(message, unicode), type(message)
399
312
self.message = message
313
self._escape_commit_message()
401
315
# Add revision data to the local branch
402
316
self.rev_id = self.builder.commit(self.message)
405
mutter("aborting commit write group because of exception:")
406
trace.log_exception_quietly()
407
note("aborting commit write group: %r" % (e,))
411
self._process_pre_hooks(old_revno, new_revno)
413
# Upload revision data to the master.
414
# this will propagate merged revisions too if needed.
415
if self.bound_branch:
416
self._set_progress_stage("Uploading data to master branch")
417
# 'commit' to the master first so a timeout here causes the
418
# local branch to be out of date
419
self.master_branch.import_last_revision_info(
420
self.branch.repository, new_revno, self.rev_id)
422
# and now do the commit locally.
423
self.branch.set_last_revision_info(new_revno, self.rev_id)
425
# Make the working tree be up to date with the branch. This
426
# includes automatic changes scheduled to be made to the tree, such
427
# as updating its basis and unversioning paths that were missing.
428
self.work_tree.unversion(self.deleted_ids)
429
self._set_progress_stage("Updating the working tree")
430
self.work_tree.update_basis_by_delta(self.rev_id,
431
self.builder.get_basis_delta())
432
self.reporter.completed(new_revno, self.rev_id)
433
self._process_post_hooks(old_revno, new_revno)
318
# Upload revision data to the master.
319
# this will propagate merged revisions too if needed.
320
if self.bound_branch:
321
self._set_progress_stage("Uploading data to master branch")
322
self.master_branch.repository.fetch(self.branch.repository,
323
revision_id=self.rev_id)
324
# now the master has the revision data
325
# 'commit' to the master first so a timeout here causes the
326
# local branch to be out of date
327
self.master_branch.set_last_revision_info(new_revno,
330
# and now do the commit locally.
331
self.branch.set_last_revision_info(new_revno, self.rev_id)
333
# Make the working tree up to date with the branch
334
self._set_progress_stage("Updating the working tree")
335
rev_tree = self.builder.revision_tree()
336
self.work_tree.set_parent_trees([(self.rev_id, rev_tree)])
337
self.reporter.completed(new_revno, self.rev_id)
338
self._process_hooks(old_revno, new_revno)
434
341
return self.rev_id
436
def _select_reporter(self):
437
"""Select the CommitReporter to use."""
439
return NullCommitReporter()
440
return ReportCommitToLog()
343
def _any_real_changes(self):
344
"""Are there real changes between new_inventory and basis?
346
For trees without rich roots, inv.root.revision changes every commit.
347
But if that is the only change, we want to treat it as though there
350
new_entries = self.builder.new_inventory.iter_entries()
351
basis_entries = self.basis_inv.iter_entries()
352
new_path, new_root_ie = new_entries.next()
353
basis_path, basis_root_ie = basis_entries.next()
355
# This is a copy of InventoryEntry.__eq__ only leaving out .revision
356
def ie_equal_no_revision(this, other):
357
return ((this.file_id == other.file_id)
358
and (this.name == other.name)
359
and (this.symlink_target == other.symlink_target)
360
and (this.text_sha1 == other.text_sha1)
361
and (this.text_size == other.text_size)
362
and (this.text_id == other.text_id)
363
and (this.parent_id == other.parent_id)
364
and (this.kind == other.kind)
365
and (this.executable == other.executable)
366
and (this.reference_revision == other.reference_revision)
368
if not ie_equal_no_revision(new_root_ie, basis_root_ie):
371
for new_ie, basis_ie in zip(new_entries, basis_entries):
372
if new_ie != basis_ie:
375
# No actual changes present
442
378
def _check_pointless(self):
443
379
if self.allow_pointless:
565
489
old_revid = self.parents[0]
567
491
old_revid = bzrlib.revision.NULL_REVISION
569
if hook_name == "pre_commit":
570
future_tree = self.builder.revision_tree()
571
tree_delta = future_tree.changes_from(self.basis_tree,
574
for hook in Branch.hooks[hook_name]:
492
for hook in Branch.hooks['post_commit']:
575
493
# show the running hook in the progress bar. As hooks may
576
494
# end up doing nothing (e.g. because they are not configured by
577
495
# the user) this is still showing progress, not showing overall
578
496
# actions - its up to each plugin to show a UI if it want's to
579
497
# (such as 'Emailing diff to foo@example.com').
580
self.pb_stage_name = "Running %s hooks [%s]" % \
581
(hook_name, Branch.hooks.get_hook_name(hook))
498
self.pb_stage_name = "Running post commit hooks [%s]" % \
499
Branch.hooks.get_hook_name(hook)
582
500
self._emit_progress()
583
501
if 'hooks' in debug.debug_flags:
584
502
mutter("Invoking commit hook: %r", hook)
585
if hook_name == "post_commit":
586
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
588
elif hook_name == "pre_commit":
589
hook(hook_local, hook_master,
590
old_revno, old_revid, new_revno, self.rev_id,
591
tree_delta, future_tree)
503
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
507
"""Cleanup any open locks, progress bars etc."""
508
cleanups = [self._cleanup_bound_branch,
509
self.basis_tree.unlock,
510
self.work_tree.unlock,
512
found_exception = None
513
for cleanup in cleanups:
516
# we want every cleanup to run no matter what.
517
# so we have a catchall here, but we will raise the
518
# last encountered exception up the stack: and
519
# typically this will be useful enough.
522
if found_exception is not None:
523
# don't do a plan raise, because the last exception may have been
524
# trashed, e is our sure-to-work exception even though it loses the
525
# full traceback. XXX: RBC 20060421 perhaps we could check the
526
# exc_info and if its the same one do a plain raise otherwise
527
# 'raise e' as we do now.
530
def _cleanup_bound_branch(self):
531
"""Executed at the end of a try/finally to cleanup a bound branch.
533
If the branch wasn't bound, this is a no-op.
534
If it was, it resents self.branch to the local branch, instead
537
if not self.bound_branch:
539
if self.master_locked:
540
self.master_branch.unlock()
542
def _escape_commit_message(self):
543
"""Replace xml-incompatible control characters."""
544
# FIXME: RBC 20060419 this should be done by the revision
545
# serialiser not by commit. Then we can also add an unescaper
546
# in the deserializer and start roundtripping revision messages
547
# precisely. See repository_implementations/test_repository.py
549
# Python strings can include characters that can't be
550
# represented in well-formed XML; escape characters that
551
# aren't listed in the XML specification
552
# (http://www.w3.org/TR/REC-xml/#NT-Char).
553
self.message, escape_count = re.subn(
554
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
555
lambda match: match.group(0).encode('unicode_escape'),
558
self.reporter.escaped(escape_count, self.message)
593
560
def _gather_parents(self):
594
561
"""Record the parents of a merge for merge detection."""
595
# TODO: Make sure that this list doesn't contain duplicate
562
# TODO: Make sure that this list doesn't contain duplicate
596
563
# entries and the order is preserved when doing this.
597
if self.use_record_iter_changes:
599
self.basis_inv = self.basis_tree.inventory
564
self.parents = self.work_tree.get_parent_ids()
600
565
self.parent_invs = [self.basis_inv]
601
566
for revision in self.parents[1:]:
602
567
if self.branch.repository.has_revision(revision):
609
574
def _update_builder_with_changes(self):
610
575
"""Update the commit builder with the data about what has changed.
612
exclude = self.exclude
577
# Build the revision inventory.
579
# This starts by creating a new empty inventory. Depending on
580
# which files are selected for commit, and what is present in the
581
# current tree, the new inventory is populated. inventory entries
582
# which are candidates for modification have their revision set to
583
# None; inventory entries that are carried over untouched have their
584
# revision set to their prior value.
586
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
587
# results to create a new inventory at the same time, which results
588
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
613
591
specific_files = self.specific_files
614
592
mutter("Selecting files for commit with filter %s", specific_files)
617
if self.use_record_iter_changes:
618
iter_changes = self.work_tree.iter_changes(self.basis_tree,
619
specific_files=specific_files)
620
iter_changes = self._filter_iter_changes(iter_changes)
621
for file_id, path, fs_hash in self.builder.record_iter_changes(
622
self.work_tree, self.basis_revid, iter_changes):
623
self.work_tree._observed_sha1(file_id, path, fs_hash)
625
# Build the new inventory
626
self._populate_from_inventory()
627
self._record_unselected()
628
self._report_and_accumulate_deletes()
630
def _filter_iter_changes(self, iter_changes):
631
"""Process iter_changes.
633
This method reports on the changes in iter_changes to the user, and
634
converts 'missing' entries in the iter_changes iterator to 'deleted'
635
entries. 'missing' entries have their
637
:param iter_changes: An iter_changes to process.
638
:return: A generator of changes.
640
reporter = self.reporter
641
report_changes = reporter.is_verbose()
643
for change in iter_changes:
645
old_path = change[1][0]
646
new_path = change[1][1]
647
versioned = change[3][1]
649
versioned = change[3][1]
650
if kind is None and versioned:
653
reporter.missing(new_path)
654
deleted_ids.append(change[0])
655
# Reset the new path (None) and new versioned flag (False)
656
change = (change[0], (change[1][0], None), change[2],
657
(change[3][0], False)) + change[4:]
658
elif kind == 'tree-reference':
659
if self.recursive == 'down':
660
self._commit_nested_tree(change[0], change[1][1])
661
if change[3][0] or change[3][1]:
665
reporter.deleted(old_path)
666
elif old_path is None:
667
reporter.snapshot_change('added', new_path)
668
elif old_path != new_path:
669
reporter.renamed('renamed', old_path, new_path)
672
self.work_tree.branch.repository._format.rich_root_data):
673
# Don't report on changes to '' in non rich root
675
reporter.snapshot_change('modified', new_path)
676
self._next_progress_entry()
677
# Unversion IDs that were found to be deleted
678
self.deleted_ids = deleted_ids
680
def _record_unselected(self):
594
# Check and warn about old CommitBuilders
595
if not self.builder.record_root_entry:
596
symbol_versioning.warn('CommitBuilders should support recording'
597
' the root entry as of bzr 0.10.', DeprecationWarning,
599
self.builder.new_inventory.add(self.basis_inv.root.copy())
601
# Build the new inventory
602
self._populate_from_inventory(specific_files)
681
604
# If specific files are selected, then all un-selected files must be
682
605
# recorded in their previous state. For more details, see
683
606
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
684
if self.specific_files or self.exclude:
685
specific_files = self.specific_files or []
686
for path, old_ie in self.basis_inv.iter_entries():
687
if old_ie.file_id in self.builder.new_inventory:
688
# already added - skip.
690
if (is_inside_any(specific_files, path)
691
and not is_inside_any(self.exclude, path)):
692
# was inside the selected path, and not excluded - if not
693
# present it has been deleted so skip.
695
# From here down it was either not selected, or was excluded:
696
# We preserve the entry unaltered.
698
# Note: specific file commits after a merge are currently
699
# prohibited. This test is for sanity/safety in case it's
700
# required after that changes.
701
if len(self.parents) > 1:
608
for path, new_ie in self.basis_inv.iter_entries():
609
if new_ie.file_id in self.builder.new_inventory:
611
if is_inside_any(specific_files, path):
703
615
self.builder.record_entry_contents(ie, self.parent_invs, path,
704
self.basis_tree, None)
706
def _report_and_accumulate_deletes(self):
707
if (isinstance(self.basis_inv, Inventory)
708
and isinstance(self.builder.new_inventory, Inventory)):
709
# the older Inventory classes provide a _byid dict, and building a
710
# set from the keys of this dict is substantially faster than even
711
# getting a set of ids from the inventory
713
# <lifeless> set(dict) is roughly the same speed as
714
# set(iter(dict)) and both are significantly slower than
716
deleted_ids = set(self.basis_inv._byid.keys()) - \
717
set(self.builder.new_inventory._byid.keys())
719
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
721
self.any_entries_deleted = True
722
deleted = [(self.basis_tree.id2path(file_id), file_id)
723
for file_id in deleted_ids]
725
# XXX: this is not quite directory-order sorting
726
for path, file_id in deleted:
727
self.builder.record_delete(path, file_id)
618
# Report what was deleted. We could skip this when no deletes are
619
# detected to gain a performance win, but it arguably serves as a
620
# 'safety check' by informing the user whenever anything disappears.
621
for path, ie in self.basis_inv.iter_entries():
622
if ie.file_id not in self.builder.new_inventory:
728
623
self.reporter.deleted(path)
730
def _check_strict(self):
731
# XXX: when we use iter_changes this would likely be faster if
732
# iter_changes would check for us (even in the presence of
625
def _populate_from_inventory(self, specific_files):
626
"""Populate the CommitBuilder by walking the working tree inventory."""
735
628
# raise an exception as soon as we find a single unknown.
736
629
for unknown in self.work_tree.unknowns():
737
630
raise StrictCommitFailed()
739
def _populate_from_inventory(self):
740
"""Populate the CommitBuilder by walking the working tree inventory."""
741
# Build the revision inventory.
743
# This starts by creating a new empty inventory. Depending on
744
# which files are selected for commit, and what is present in the
745
# current tree, the new inventory is populated. inventory entries
746
# which are candidates for modification have their revision set to
747
# None; inventory entries that are carried over untouched have their
748
# revision set to their prior value.
750
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
751
# results to create a new inventory at the same time, which results
752
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
755
specific_files = self.specific_files
756
exclude = self.exclude
757
report_changes = self.reporter.is_verbose()
759
# A tree of paths that have been deleted. E.g. if foo/bar has been
760
# deleted, then we have {'foo':{'bar':{}}}
762
# XXX: Note that entries may have the wrong kind because the entry does
763
# not reflect the status on disk.
633
deleted_paths = set()
764
634
work_inv = self.work_tree.inventory
765
# NB: entries will include entries within the excluded ids/paths
766
# because iter_entries_by_dir has no 'exclude' facility today.
767
entries = work_inv.iter_entries_by_dir(
768
specific_file_ids=self.specific_file_ids, yield_parents=True)
635
assert work_inv.root is not None
636
entries = work_inv.iter_entries()
637
if not self.builder.record_root_entry:
769
639
for path, existing_ie in entries:
770
640
file_id = existing_ie.file_id
771
641
name = existing_ie.name
772
642
parent_id = existing_ie.parent_id
773
643
kind = existing_ie.kind
644
if kind == 'directory':
645
self._next_progress_entry()
774
647
# Skip files that have been deleted from the working tree.
775
# The deleted path ids are also recorded so they can be explicitly
778
path_segments = splitpath(path)
779
deleted_dict = deleted_paths
780
for segment in path_segments:
781
deleted_dict = deleted_dict.get(segment, None)
783
# We either took a path not present in the dict
784
# (deleted_dict was None), or we've reached an empty
785
# child dir in the dict, so are now a sub-path.
789
if deleted_dict is not None:
790
# the path has a deleted parent, do not add it.
792
if exclude and is_inside_any(exclude, path):
793
# Skip excluded paths. Excluded paths are processed by
794
# _update_builder_with_changes.
648
# The deleted files/directories are also recorded so they
649
# can be explicitly unversioned later. Note that when a
650
# filter of specific files is given, we must only skip/record
651
# deleted files matching that filter.
652
if is_inside_any(deleted_paths, path):
796
content_summary = self.work_tree.path_content_summary(path)
797
kind = content_summary[0]
798
# Note that when a filter of specific files is given, we must only
799
# skip/record deleted files matching that filter.
800
654
if not specific_files or is_inside_any(specific_files, path):
801
if kind == 'missing':
802
if not deleted_paths:
803
# path won't have been split yet.
804
path_segments = splitpath(path)
805
deleted_dict = deleted_paths
806
for segment in path_segments:
807
deleted_dict = deleted_dict.setdefault(segment, {})
655
if not self.work_tree.has_filename(path):
656
deleted_paths.add(path)
808
657
self.reporter.missing(path)
809
self._next_progress_entry()
810
658
deleted_ids.append(file_id)
812
# TODO: have the builder do the nested commit just-in-time IF and
814
if kind == 'tree-reference':
815
# enforce repository nested tree policy.
816
if (not self.work_tree.supports_tree_reference() or
817
# repository does not support it either.
818
not self.branch.repository._format.supports_tree_reference):
820
content_summary = (kind, None, None, None)
821
elif self.recursive == 'down':
822
nested_revision_id = self._commit_nested_tree(
824
content_summary = (kind, None, None, nested_revision_id)
826
nested_revision_id = self.work_tree.get_reference_revision(file_id)
827
content_summary = (kind, None, None, nested_revision_id)
661
kind = self.work_tree.kind(file_id)
662
# TODO: specific_files filtering before nested tree processing
663
if kind == 'tree-reference' and self.recursive == 'down':
664
self._commit_nested_tree(file_id, path)
665
except errors.NoSuchFile:
829
668
# Record an entry for this item
830
669
# Note: I don't particularly want to have the existing_ie
831
670
# parameter but the test suite currently (28-Jun-07) breaks
832
671
# without it thanks to a unicode normalisation issue. :-(
833
definitely_changed = kind != existing_ie.kind
672
definitely_changed = kind != existing_ie.kind
834
673
self._record_entry(path, file_id, specific_files, kind, name,
835
parent_id, definitely_changed, existing_ie, report_changes,
674
parent_id, definitely_changed, existing_ie)
838
676
# Unversion IDs that were found to be deleted
839
self.deleted_ids = deleted_ids
677
self.work_tree.unversion(deleted_ids)
841
679
def _commit_nested_tree(self, file_id, path):
842
680
"Commit a nested tree."
861
700
strict=self.strict, verbose=self.verbose,
862
701
local=self.local, reporter=self.reporter)
863
702
except errors.PointlessCommit:
864
return self.work_tree.get_reference_revision(file_id)
866
705
def _record_entry(self, path, file_id, specific_files, kind, name,
867
parent_id, definitely_changed, existing_ie, report_changes,
706
parent_id, definitely_changed, existing_ie=None):
869
707
"Record the new inventory entry for a path if any."
870
708
# mutter('check %s {%s}', path, file_id)
871
# mutter('%s selected for commit', path)
872
if definitely_changed or existing_ie is None:
873
ie = make_entry(kind, name, parent_id, file_id)
709
if (not specific_files or
710
is_inside_or_parent_of_any(specific_files, path)):
711
# mutter('%s selected for commit', path)
712
if definitely_changed or existing_ie is None:
713
ie = inventory.make_entry(kind, name, parent_id, file_id)
715
ie = existing_ie.copy()
875
ie = existing_ie.copy()
877
# For carried over entries we don't care about the fs hash - the repo
878
# isn't generating a sha, so we're not saving computation time.
879
_, _, fs_hash = self.builder.record_entry_contents(
880
ie, self.parent_invs, path, self.work_tree, content_summary)
718
# mutter('%s not selected for commit', path)
719
if self.basis_inv.has_id(file_id):
720
ie = self.basis_inv[file_id].copy()
722
# this entry is new and not being committed
725
self.builder.record_entry_contents(ie, self.parent_invs,
726
path, self.work_tree)
882
727
self._report_change(ie, path)
884
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
887
730
def _report_change(self, ie, path):
added
removed
bzrlib/commit.py
