136
179
working inventory.
138
181
def __init__(self,
140
if reporter is not None:
141
self.reporter = reporter
143
self.reporter = NullCommitReporter()
184
"""Create a Commit object.
186
:param reporter: the default reporter to use or None to decide later
188
self.reporter = reporter
151
196
specific_files=None,
153
198
allow_pointless=True,
206
message_callback=None,
209
possible_master_transports=None):
155
210
"""Commit working copy as a new revision.
157
timestamp -- if not None, seconds-since-epoch for a
158
postdated/predated commit.
160
specific_files -- If true, commit only those files.
162
rev_id -- If set, use this as the new revision id.
212
:param message: the commit message (it or message_callback is required)
214
:param timestamp: if not None, seconds-since-epoch for a
215
postdated/predated commit.
217
:param specific_files: If true, commit only those files.
219
:param rev_id: If set, use this as the new revision id.
163
220
Useful for test or import commands that need to tightly
164
221
control what revisions are assigned. If you duplicate
165
222
a revision id that exists elsewhere it is your own fault.
166
223
If null (default), a time/random revision id is generated.
168
allow_pointless -- If true (default), commit even if nothing
225
:param allow_pointless: If true (default), commit even if nothing
169
226
has changed and no merges are recorded.
228
:param strict: If true, don't allow a commit if the working tree
229
contains unknown files.
231
:param revprops: Properties for new revision
232
:param local: Perform a local only commit.
233
:param reporter: the reporter to use or None for the default
234
:param verbose: if True and the reporter is not None, report everything
235
:param recursive: If set to 'down', commit in any subtrees that have
236
pending changes of any sort during this commit.
237
:param exclude: None or a list of relative paths to exclude from the
238
commit. Pending changes to excluded files will be ignored by the
171
241
mutter('preparing to commit')
174
self.weave_store = branch.weave_store
176
self.specific_files = specific_files
243
if working_tree is None:
244
raise BzrError("working_tree must be passed into commit().")
246
self.work_tree = working_tree
247
self.branch = self.work_tree.branch
248
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
249
if not self.branch.repository.supports_rich_root():
250
raise errors.RootNotRich()
251
if message_callback is None:
252
if message is not None:
253
if isinstance(message, str):
254
message = message.decode(get_user_encoding())
255
message_callback = lambda x: message
257
raise BzrError("The message or message_callback keyword"
258
" parameter is required for commit().")
260
self.bound_branch = None
261
self.any_entries_deleted = False
262
if exclude is not None:
263
self.exclude = sorted(
264
minimum_path_selection(exclude))
268
self.master_branch = None
269
self.master_locked = False
270
self.recursive = recursive
272
if specific_files is not None:
273
self.specific_files = sorted(
274
minimum_path_selection(specific_files))
276
self.specific_files = None
177
278
self.allow_pointless = allow_pointless
179
if timestamp is None:
180
self.timestamp = time.time()
182
self.timestamp = long(timestamp)
185
self.rev_id = _gen_revision_id(self.branch, self.timestamp)
189
if committer is None:
190
self.committer = username(self.branch)
192
assert isinstance(committer, basestring), type(committer)
193
self.committer = committer
196
self.timezone = local_time_offset()
198
self.timezone = int(timezone)
200
assert isinstance(message, basestring), type(message)
201
self.message = message
202
self._escape_commit_message()
204
self.branch.lock_write()
279
self.revprops = revprops
280
self.message_callback = message_callback
281
self.timestamp = timestamp
282
self.timezone = timezone
283
self.committer = committer
285
self.verbose = verbose
287
self.work_tree.lock_write()
288
self.parents = self.work_tree.get_parent_ids()
289
# We can use record_iter_changes IFF iter_changes is compatible with
290
# the command line parameters, and the repository has fast delta
291
# generation. See bug 347649.
292
self.use_record_iter_changes = (
293
not self.specific_files and
295
not self.branch.repository._format.supports_tree_reference and
296
(self.branch.repository._format.fast_deltas or
297
len(self.parents) < 2))
298
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
299
self.basis_revid = self.work_tree.last_revision()
300
self.basis_tree = self.work_tree.basis_tree()
301
self.basis_tree.lock_read()
206
self.work_tree = self.branch.working_tree()
207
self.work_inv = self.work_tree.inventory
208
self.basis_tree = self.branch.basis_tree()
209
self.basis_inv = self.basis_tree.inventory
303
# Cannot commit with conflicts present.
304
if len(self.work_tree.conflicts()) > 0:
305
raise ConflictsInTree
307
# Setup the bound branch variables as needed.
308
self._check_bound_branch(possible_master_transports)
310
# Check that the working tree is up to date
311
old_revno, new_revno = self._check_out_of_date_tree()
313
# Complete configuration setup
314
if reporter is not None:
315
self.reporter = reporter
316
elif self.reporter is None:
317
self.reporter = self._select_reporter()
318
if self.config is None:
319
self.config = self.branch.get_config()
321
self._set_specific_file_ids()
323
# Setup the progress bar. As the number of files that need to be
324
# committed in unknown, progress is reported as stages.
325
# We keep track of entries separately though and include that
326
# information in the progress bar during the relevant stages.
327
self.pb_stage_name = ""
328
self.pb_stage_count = 0
329
self.pb_stage_total = 5
330
if self.bound_branch:
331
self.pb_stage_total += 1
332
self.pb.show_pct = False
333
self.pb.show_spinner = False
334
self.pb.show_eta = False
335
self.pb.show_count = True
336
self.pb.show_bar = True
211
338
self._gather_parents()
339
# After a merge, a selected file commit is not supported.
340
# See 'bzr help merge' for an explanation as to why.
212
341
if len(self.parents) > 1 and self.specific_files:
213
raise NotImplementedError('selected-file commit of merges is not supported yet')
214
self._check_parents_present()
216
self._remove_deleted()
217
self.new_inv = Inventory()
218
self._store_entries()
219
self._report_deletes()
220
self._set_name_versions()
222
if not (self.allow_pointless
223
or len(self.parents) > 1
224
or self.new_inv != self.basis_inv):
225
raise PointlessCommit()
227
self._record_inventory()
228
self._record_ancestry()
229
self._make_revision()
230
note('committed r%d {%s}', (self.branch.revno() + 1),
232
self.branch.append_revision(self.rev_id)
233
self.branch.set_pending_merges([])
342
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
343
# Excludes are a form of selected file commit.
344
if len(self.parents) > 1 and self.exclude:
345
raise errors.CannotCommitSelectedFileMerge(self.exclude)
347
# Collect the changes
348
self._set_progress_stage("Collecting changes", counter=True)
349
self.builder = self.branch.get_commit_builder(self.parents,
350
self.config, timestamp, timezone, committer, revprops, rev_id)
353
self.builder.will_record_deletes()
354
# find the location being committed to
355
if self.bound_branch:
356
master_location = self.master_branch.base
358
master_location = self.branch.base
360
# report the start of the commit
361
self.reporter.started(new_revno, self.rev_id, master_location)
363
self._update_builder_with_changes()
364
self._check_pointless()
366
# TODO: Now the new inventory is known, check for conflicts.
367
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
368
# weave lines, because nothing should be recorded until it is known
369
# that commit will succeed.
370
self._set_progress_stage("Saving data locally")
371
self.builder.finish_inventory()
373
# Prompt the user for a commit message if none provided
374
message = message_callback(self)
375
self.message = message
376
self._escape_commit_message()
378
# Add revision data to the local branch
379
self.rev_id = self.builder.commit(self.message)
382
mutter("aborting commit write group because of exception:")
383
trace.log_exception_quietly()
384
note("aborting commit write group: %r" % (e,))
388
self._process_pre_hooks(old_revno, new_revno)
390
# Upload revision data to the master.
391
# this will propagate merged revisions too if needed.
392
if self.bound_branch:
393
self._set_progress_stage("Uploading data to master branch")
394
# 'commit' to the master first so a timeout here causes the
395
# local branch to be out of date
396
self.master_branch.import_last_revision_info(
397
self.branch.repository, new_revno, self.rev_id)
399
# and now do the commit locally.
400
self.branch.set_last_revision_info(new_revno, self.rev_id)
402
# Make the working tree up to date with the branch
403
self._set_progress_stage("Updating the working tree")
404
self.work_tree.update_basis_by_delta(self.rev_id,
405
self.builder.get_basis_delta())
406
self.reporter.completed(new_revno, self.rev_id)
407
self._process_post_hooks(old_revno, new_revno)
239
def _record_inventory(self):
240
"""Store the inventory for the new revision."""
241
inv_text = serializer_v5.write_inventory_to_string(self.new_inv)
242
self.inv_sha1 = sha_string(inv_text)
243
s = self.branch.control_weaves
244
s.add_text('inventory', self.rev_id,
245
split_lines(inv_text), self.parents)
412
def _select_reporter(self):
413
"""Select the CommitReporter to use."""
415
return NullCommitReporter()
416
return ReportCommitToLog()
418
def _check_pointless(self):
419
if self.allow_pointless:
421
# A merge with no effect on files
422
if len(self.parents) > 1:
424
# TODO: we could simplify this by using self.builder.basis_delta.
426
# The initial commit adds a root directory, but this in itself is not
427
# a worthwhile commit.
428
if (self.basis_revid == revision.NULL_REVISION and
429
((self.builder.new_inventory is not None and
430
len(self.builder.new_inventory) == 1) or
431
len(self.builder._basis_delta) == 1)):
432
raise PointlessCommit()
433
if self.builder.any_changes():
435
raise PointlessCommit()
437
def _check_bound_branch(self, possible_master_transports=None):
438
"""Check to see if the local branch is bound.
440
If it is bound, then most of the commit will actually be
441
done using the remote branch as the target branch.
442
Only at the end will the local branch be updated.
444
if self.local and not self.branch.get_bound_location():
445
raise errors.LocalRequiresBoundBranch()
448
self.master_branch = self.branch.get_master_branch(
449
possible_master_transports)
451
if not self.master_branch:
452
# make this branch the reference branch for out of date checks.
453
self.master_branch = self.branch
456
# If the master branch is bound, we must fail
457
master_bound_location = self.master_branch.get_bound_location()
458
if master_bound_location:
459
raise errors.CommitToDoubleBoundBranch(self.branch,
460
self.master_branch, master_bound_location)
462
# TODO: jam 20051230 We could automatically push local
463
# commits to the remote branch if they would fit.
464
# But for now, just require remote to be identical
467
# Make sure the local branch is identical to the master
468
master_info = self.master_branch.last_revision_info()
469
local_info = self.branch.last_revision_info()
470
if local_info != master_info:
471
raise errors.BoundBranchOutOfDate(self.branch,
474
# Now things are ready to change the master branch
476
self.bound_branch = self.branch
477
self.master_branch.lock_write()
478
self.master_locked = True
480
def _check_out_of_date_tree(self):
481
"""Check that the working tree is up to date.
483
:return: old_revision_number,new_revision_number tuple
486
first_tree_parent = self.work_tree.get_parent_ids()[0]
488
# if there are no parents, treat our parent as 'None'
489
# this is so that we still consider the master branch
490
# - in a checkout scenario the tree may have no
491
# parents but the branch may do.
492
first_tree_parent = bzrlib.revision.NULL_REVISION
493
old_revno, master_last = self.master_branch.last_revision_info()
494
if master_last != first_tree_parent:
495
if master_last != bzrlib.revision.NULL_REVISION:
496
raise errors.OutOfDateTree(self.work_tree)
497
if self.branch.repository.has_revision(first_tree_parent):
498
new_revno = old_revno + 1
500
# ghost parents never appear in revision history.
502
return old_revno,new_revno
504
def _process_pre_hooks(self, old_revno, new_revno):
505
"""Process any registered pre commit hooks."""
506
self._set_progress_stage("Running pre_commit hooks")
507
self._process_hooks("pre_commit", old_revno, new_revno)
509
def _process_post_hooks(self, old_revno, new_revno):
510
"""Process any registered post commit hooks."""
511
# Process the post commit hooks, if any
512
self._set_progress_stage("Running post_commit hooks")
513
# old style commit hooks - should be deprecated ? (obsoleted in
515
if self.config.post_commit() is not None:
516
hooks = self.config.post_commit().split(' ')
517
# this would be nicer with twisted.python.reflect.namedAny
519
result = eval(hook + '(branch, rev_id)',
520
{'branch':self.branch,
522
'rev_id':self.rev_id})
523
# process new style post commit hooks
524
self._process_hooks("post_commit", old_revno, new_revno)
526
def _process_hooks(self, hook_name, old_revno, new_revno):
527
if not Branch.hooks[hook_name]:
530
# new style commit hooks:
531
if not self.bound_branch:
532
hook_master = self.branch
535
hook_master = self.master_branch
536
hook_local = self.branch
537
# With bound branches, when the master is behind the local branch,
538
# the 'old_revno' and old_revid values here are incorrect.
539
# XXX: FIXME ^. RBC 20060206
541
old_revid = self.parents[0]
543
old_revid = bzrlib.revision.NULL_REVISION
545
if hook_name == "pre_commit":
546
future_tree = self.builder.revision_tree()
547
tree_delta = future_tree.changes_from(self.basis_tree,
550
for hook in Branch.hooks[hook_name]:
551
# show the running hook in the progress bar. As hooks may
552
# end up doing nothing (e.g. because they are not configured by
553
# the user) this is still showing progress, not showing overall
554
# actions - its up to each plugin to show a UI if it want's to
555
# (such as 'Emailing diff to foo@example.com').
556
self.pb_stage_name = "Running %s hooks [%s]" % \
557
(hook_name, Branch.hooks.get_hook_name(hook))
558
self._emit_progress()
559
if 'hooks' in debug.debug_flags:
560
mutter("Invoking commit hook: %r", hook)
561
if hook_name == "post_commit":
562
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
564
elif hook_name == "pre_commit":
565
hook(hook_local, hook_master,
566
old_revno, old_revid, new_revno, self.rev_id,
567
tree_delta, future_tree)
570
"""Cleanup any open locks, progress bars etc."""
571
cleanups = [self._cleanup_bound_branch,
572
self.basis_tree.unlock,
573
self.work_tree.unlock,
575
found_exception = None
576
for cleanup in cleanups:
579
# we want every cleanup to run no matter what.
580
# so we have a catchall here, but we will raise the
581
# last encountered exception up the stack: and
582
# typically this will be useful enough.
585
if found_exception is not None:
586
# don't do a plan raise, because the last exception may have been
587
# trashed, e is our sure-to-work exception even though it loses the
588
# full traceback. XXX: RBC 20060421 perhaps we could check the
589
# exc_info and if its the same one do a plain raise otherwise
590
# 'raise e' as we do now.
593
def _cleanup_bound_branch(self):
594
"""Executed at the end of a try/finally to cleanup a bound branch.
596
If the branch wasn't bound, this is a no-op.
597
If it was, it resents self.branch to the local branch, instead
600
if not self.bound_branch:
602
if self.master_locked:
603
self.master_branch.unlock()
247
605
def _escape_commit_message(self):
248
606
"""Replace xml-incompatible control characters."""
249
# Python strings can include characters that can't be
250
# represented in well-formed XML; escape characters that
251
# aren't listed in the XML specification
252
# (http://www.w3.org/TR/REC-xml/#NT-Char).
253
if isinstance(self.message, unicode):
254
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
256
# Use a regular 'str' as pattern to avoid having re.subn
257
# return 'unicode' results.
258
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
259
self.message, escape_count = re.subn(
261
lambda match: match.group(0).encode('unicode_escape'),
607
# FIXME: RBC 20060419 this should be done by the revision
608
# serialiser not by commit. Then we can also add an unescaper
609
# in the deserializer and start roundtripping revision messages
610
# precisely. See repository_implementations/test_repository.py
611
self.message, escape_count = xml_serializer.escape_invalid_chars(
264
note("replaced %d control characters in message", escape_count)
266
def _record_ancestry(self):
267
"""Append merged revision ancestry to the ancestry file.
269
This should be the merged ancestry of all parents, plus the
271
s = self.branch.control_weaves
272
w = s.get_weave_or_empty('ancestry')
273
lines = self._make_ancestry(w)
274
w.add(self.rev_id, self.parents, lines)
275
s.put_weave('ancestry', w)
278
def _make_ancestry(self, ancestry_weave):
279
"""Return merged ancestry lines.
281
The lines are revision-ids followed by newlines."""
282
parent_ancestries = [ancestry_weave.get(p) for p in self.parents]
283
new_lines = merge_ancestry_lines(self.rev_id, parent_ancestries)
284
mutter('merged ancestry of {%s}:\n%s', self.rev_id, ''.join(new_lines))
614
self.reporter.escaped(escape_count, self.message)
288
616
def _gather_parents(self):
289
pending_merges = self.branch.pending_merges()
291
self.parent_trees = []
292
precursor_id = self.branch.last_revision()
294
self.parents.append(precursor_id)
295
self.parent_trees.append(self.basis_tree)
296
self.parents += pending_merges
297
self.parent_trees.extend(map(self.branch.revision_tree, pending_merges))
300
def _check_parents_present(self):
301
for parent_id in self.parents:
302
mutter('commit parent revision {%s}', parent_id)
303
if not self.branch.has_revision(parent_id):
304
warning("can't commit a merge from an absent parent")
305
raise HistoryMissing(self.branch, 'revision', parent_id)
308
def _make_revision(self):
309
"""Record a new revision object for this commit."""
310
self.rev = Revision(timestamp=self.timestamp,
311
timezone=self.timezone,
312
committer=self.committer,
313
message=self.message,
314
inventory_sha1=self.inv_sha1,
315
revision_id=self.rev_id)
316
self.rev.parent_ids = self.parents
318
serializer_v5.write_revision(self.rev, rev_tmp)
320
self.branch.revision_store.add(rev_tmp, self.rev_id, compressed=False)
321
mutter('new revision_id is {%s}', self.rev_id)
324
def _remove_deleted(self):
325
"""Remove deleted files from the working inventories.
327
This is done prior to taking the working inventory as the
328
basis for the new committed inventory.
330
This returns true if any files
331
*that existed in the basis inventory* were deleted.
332
Files that were added and deleted
333
in the working copy don't matter.
335
specific = self.specific_files
617
"""Record the parents of a merge for merge detection."""
618
# TODO: Make sure that this list doesn't contain duplicate
619
# entries and the order is preserved when doing this.
620
if self.use_record_iter_changes:
622
self.basis_inv = self.basis_tree.inventory
623
self.parent_invs = [self.basis_inv]
624
for revision in self.parents[1:]:
625
if self.branch.repository.has_revision(revision):
626
mutter('commit parent revision {%s}', revision)
627
inventory = self.branch.repository.get_inventory(revision)
628
self.parent_invs.append(inventory)
630
mutter('commit parent ghost revision {%s}', revision)
632
def _update_builder_with_changes(self):
633
"""Update the commit builder with the data about what has changed.
635
exclude = self.exclude
636
specific_files = self.specific_files or []
637
mutter("Selecting files for commit with filter %s", specific_files)
640
if self.use_record_iter_changes:
641
iter_changes = self.work_tree.iter_changes(self.basis_tree)
642
iter_changes = self._filter_iter_changes(iter_changes)
643
for file_id, path, fs_hash in self.builder.record_iter_changes(
644
self.work_tree, self.basis_revid, iter_changes):
645
self.work_tree._observed_sha1(file_id, path, fs_hash)
647
# Build the new inventory
648
self._populate_from_inventory()
649
self._record_unselected()
650
self._report_and_accumulate_deletes()
652
def _filter_iter_changes(self, iter_changes):
653
"""Process iter_changes.
655
This method reports on the changes in iter_changes to the user, and
656
converts 'missing' entries in the iter_changes iterator to 'deleted'
657
entries. 'missing' entries have their
659
:param iter_changes: An iter_changes to process.
660
:return: A generator of changes.
662
reporter = self.reporter
663
report_changes = reporter.is_verbose()
337
for path, ie in self.work_inv.iter_entries():
338
if specific and not is_inside_any(specific, path):
340
if not self.work_tree.has_filename(path):
341
note('missing %s', path)
342
deleted_ids.append(ie.file_id)
665
for change in iter_changes:
667
old_path = change[1][0]
668
new_path = change[1][1]
669
versioned = change[3][1]
671
versioned = change[3][1]
672
if kind is None and versioned:
675
reporter.missing(new_path)
676
deleted_ids.append(change[0])
677
# Reset the new path (None) and new versioned flag (False)
678
change = (change[0], (change[1][0], None), change[2],
679
(change[3][0], False)) + change[4:]
680
elif kind == 'tree-reference':
681
if self.recursive == 'down':
682
self._commit_nested_tree(change[0], change[1][1])
683
if change[3][0] or change[3][1]:
687
reporter.deleted(old_path)
688
elif old_path is None:
689
reporter.snapshot_change('added', new_path)
690
elif old_path != new_path:
691
reporter.renamed('renamed', old_path, new_path)
694
self.work_tree.branch.repository._format.rich_root_data):
695
# Don't report on changes to '' in non rich root
697
reporter.snapshot_change('modified', new_path)
698
self._next_progress_entry()
699
# Unversion IDs that were found to be deleted
700
self.work_tree.unversion(deleted_ids)
702
def _record_unselected(self):
703
# If specific files are selected, then all un-selected files must be
704
# recorded in their previous state. For more details, see
705
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
706
if self.specific_files or self.exclude:
707
specific_files = self.specific_files or []
708
for path, old_ie in self.basis_inv.iter_entries():
709
if old_ie.file_id in self.builder.new_inventory:
710
# already added - skip.
712
if (is_inside_any(specific_files, path)
713
and not is_inside_any(self.exclude, path)):
714
# was inside the selected path, and not excluded - if not
715
# present it has been deleted so skip.
717
# From here down it was either not selected, or was excluded:
718
# We preserve the entry unaltered.
720
# Note: specific file commits after a merge are currently
721
# prohibited. This test is for sanity/safety in case it's
722
# required after that changes.
723
if len(self.parents) > 1:
725
self.builder.record_entry_contents(ie, self.parent_invs, path,
726
self.basis_tree, None)
728
def _report_and_accumulate_deletes(self):
729
if (isinstance(self.basis_inv, Inventory)
730
and isinstance(self.builder.new_inventory, Inventory)):
731
# the older Inventory classes provide a _byid dict, and building a
732
# set from the keys of this dict is substantially faster than even
733
# getting a set of ids from the inventory
735
# <lifeless> set(dict) is roughly the same speed as
736
# set(iter(dict)) and both are significantly slower than
738
deleted_ids = set(self.basis_inv._byid.keys()) - \
739
set(self.builder.new_inventory._byid.keys())
741
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
344
for file_id in deleted_ids:
345
del self.work_inv[file_id]
346
self.branch._write_inventory(self.work_inv)
349
def _find_file_parents(self, file_id):
350
"""Return the text versions and hashes for all file parents.
352
Returned as a map from text version to inventory entry.
354
This is a set containing the file versions in all parents
355
revisions containing the file. If the file is new, the set
358
for tree in self.parent_trees:
359
if file_id in tree.inventory:
360
ie = tree.inventory[file_id]
361
assert ie.kind == 'file'
362
assert ie.file_id == file_id
363
if ie.text_version in r:
364
assert r[ie.text_version] == ie
366
r[ie.text_version] = ie
370
def _set_name_versions(self):
371
"""Pass over inventory and mark new entry version as needed.
373
Files get a new name version when they are new, have a
374
different parent, or a different name from in the
375
basis inventory, or if the file is in a different place
376
to any of the parents."""
377
# XXX: Need to think more here about when the user has
378
# made a specific decision on a particular value -- c.f.
380
for path, ie in self.new_inv.iter_entries():
383
for parent_tree in self.parent_trees:
384
parent_inv = parent_tree.inventory
385
if file_id not in parent_inv:
387
parent_ie = parent_inv[file_id]
388
if parent_ie.parent_id != ie.parent_id:
391
elif parent_ie.name != ie.name:
394
elif old_version is None:
395
old_version = parent_ie.name_version
396
elif old_version != parent_ie.name_version:
400
pass # so far so good
401
if old_version is None:
402
mutter('new name_version for {%s}', file_id)
403
ie.name_version = self.rev_id
405
mutter('name_version for {%s} inherited as {%s}',
406
file_id, old_version)
407
ie.name_version = old_version
410
def _store_entries(self):
411
"""Build revision inventory and store modified files.
413
This is called with new_inv a new empty inventory. Depending on
414
which files are selected for commit, and which ones have
415
been modified or merged, new inventory entries are built
416
based on the working and parent inventories.
418
As a side-effect this stores new text versions for committed
419
files with text changes or merges.
421
Each entry can have one of several things happen:
423
carry_file -- carried from the previous version (if not
426
commit_nonfile -- no text to worry about
428
commit_old_text -- same text, may have moved
430
commit_file -- new text version
743
self.any_entries_deleted = True
744
deleted = [(self.basis_tree.id2path(file_id), file_id)
745
for file_id in deleted_ids]
747
# XXX: this is not quite directory-order sorting
748
for path, file_id in deleted:
749
self.builder.record_delete(path, file_id)
750
self.reporter.deleted(path)
752
def _check_strict(self):
753
# XXX: when we use iter_changes this would likely be faster if
754
# iter_changes would check for us (even in the presence of
757
# raise an exception as soon as we find a single unknown.
758
for unknown in self.work_tree.unknowns():
759
raise StrictCommitFailed()
761
def _populate_from_inventory(self):
762
"""Populate the CommitBuilder by walking the working tree inventory."""
763
# Build the revision inventory.
765
# This starts by creating a new empty inventory. Depending on
766
# which files are selected for commit, and what is present in the
767
# current tree, the new inventory is populated. inventory entries
768
# which are candidates for modification have their revision set to
769
# None; inventory entries that are carried over untouched have their
770
# revision set to their prior value.
772
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
773
# results to create a new inventory at the same time, which results
774
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
777
specific_files = self.specific_files
778
exclude = self.exclude
779
report_changes = self.reporter.is_verbose()
781
# A tree of paths that have been deleted. E.g. if foo/bar has been
782
# deleted, then we have {'foo':{'bar':{}}}
784
# XXX: Note that entries may have the wrong kind because the entry does
785
# not reflect the status on disk.
786
work_inv = self.work_tree.inventory
787
# NB: entries will include entries within the excluded ids/paths
788
# because iter_entries_by_dir has no 'exclude' facility today.
789
entries = work_inv.iter_entries_by_dir(
790
specific_file_ids=self.specific_file_ids, yield_parents=True)
791
for path, existing_ie in entries:
792
file_id = existing_ie.file_id
793
name = existing_ie.name
794
parent_id = existing_ie.parent_id
795
kind = existing_ie.kind
796
# Skip files that have been deleted from the working tree.
797
# The deleted path ids are also recorded so they can be explicitly
800
path_segments = splitpath(path)
801
deleted_dict = deleted_paths
802
for segment in path_segments:
803
deleted_dict = deleted_dict.get(segment, None)
805
# We either took a path not present in the dict
806
# (deleted_dict was None), or we've reached an empty
807
# child dir in the dict, so are now a sub-path.
811
if deleted_dict is not None:
812
# the path has a deleted parent, do not add it.
814
if exclude and is_inside_any(exclude, path):
815
# Skip excluded paths. Excluded paths are processed by
816
# _update_builder_with_changes.
818
content_summary = self.work_tree.path_content_summary(path)
819
# Note that when a filter of specific files is given, we must only
820
# skip/record deleted files matching that filter.
821
if not specific_files or is_inside_any(specific_files, path):
822
if content_summary[0] == 'missing':
823
if not deleted_paths:
824
# path won't have been split yet.
825
path_segments = splitpath(path)
826
deleted_dict = deleted_paths
827
for segment in path_segments:
828
deleted_dict = deleted_dict.setdefault(segment, {})
829
self.reporter.missing(path)
830
self._next_progress_entry()
831
deleted_ids.append(file_id)
833
# TODO: have the builder do the nested commit just-in-time IF and
835
if content_summary[0] == 'tree-reference':
836
# enforce repository nested tree policy.
837
if (not self.work_tree.supports_tree_reference() or
838
# repository does not support it either.
839
not self.branch.repository._format.supports_tree_reference):
840
content_summary = ('directory',) + content_summary[1:]
841
kind = content_summary[0]
842
# TODO: specific_files filtering before nested tree processing
843
if kind == 'tree-reference':
844
if self.recursive == 'down':
845
nested_revision_id = self._commit_nested_tree(
847
content_summary = content_summary[:3] + (
850
content_summary = content_summary[:3] + (
851
self.work_tree.get_reference_revision(file_id),)
853
# Record an entry for this item
854
# Note: I don't particularly want to have the existing_ie
855
# parameter but the test suite currently (28-Jun-07) breaks
856
# without it thanks to a unicode normalisation issue. :-(
857
definitely_changed = kind != existing_ie.kind
858
self._record_entry(path, file_id, specific_files, kind, name,
859
parent_id, definitely_changed, existing_ie, report_changes,
862
# Unversion IDs that were found to be deleted
863
self.work_tree.unversion(deleted_ids)
865
def _commit_nested_tree(self, file_id, path):
866
"Commit a nested tree."
867
sub_tree = self.work_tree.get_nested_tree(file_id, path)
868
# FIXME: be more comprehensive here:
869
# this works when both trees are in --trees repository,
870
# but when both are bound to a different repository,
871
# it fails; a better way of approaching this is to
872
# finally implement the explicit-caches approach design
873
# a while back - RBC 20070306.
874
if sub_tree.branch.repository.has_same_location(
875
self.work_tree.branch.repository):
876
sub_tree.branch.repository = \
877
self.work_tree.branch.repository
879
return sub_tree.commit(message=None, revprops=self.revprops,
880
recursive=self.recursive,
881
message_callback=self.message_callback,
882
timestamp=self.timestamp, timezone=self.timezone,
883
committer=self.committer,
884
allow_pointless=self.allow_pointless,
885
strict=self.strict, verbose=self.verbose,
886
local=self.local, reporter=self.reporter)
887
except errors.PointlessCommit:
888
return self.work_tree.get_reference_revision(file_id)
890
def _record_entry(self, path, file_id, specific_files, kind, name,
891
parent_id, definitely_changed, existing_ie, report_changes,
893
"Record the new inventory entry for a path if any."
894
# mutter('check %s {%s}', path, file_id)
895
# mutter('%s selected for commit', path)
896
if definitely_changed or existing_ie is None:
897
ie = make_entry(kind, name, parent_id, file_id)
899
ie = existing_ie.copy()
901
# For carried over entries we don't care about the fs hash - the repo
902
# isn't generating a sha, so we're not saving computation time.
903
_, _, fs_hash = self.builder.record_entry_contents(
904
ie, self.parent_invs, path, self.work_tree, content_summary)
906
self._report_change(ie, path)
908
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
911
def _report_change(self, ie, path):
912
"""Report a change to the user.
914
The change that has occurred is described relative to the basis
432
for path, new_ie in self.work_inv.iter_entries():
433
file_id = new_ie.file_id
434
mutter('check %s {%s}', path, new_ie.file_id)
435
if self.specific_files:
436
if not is_inside_any(self.specific_files, path):
437
mutter('%s not selected for commit', path)
438
self._carry_file(file_id)
440
if new_ie.kind != 'file':
441
self._commit_nonfile(file_id)
444
file_parents = self._find_file_parents(file_id)
445
mutter('parents of %s are %r', path, file_parents)
446
if len(file_parents) == 1:
447
parent_ie = file_parents.values()[0]
448
wc_sha1 = self.work_tree.get_file_sha1(file_id)
449
if parent_ie.text_sha1 == wc_sha1:
450
# text not changed or merged
451
self._commit_old_text(file_id, parent_ie)
453
# file is either new, or a file merge; need to record
455
if len(file_parents) > 1:
456
note('merged %s', path)
457
elif len(file_parents) == 0:
458
note('added %s', path)
917
if (self.basis_inv.has_id(ie.file_id)):
918
basis_ie = self.basis_inv[ie.file_id]
921
change = ie.describe_change(basis_ie, ie)
922
if change in (InventoryEntry.RENAMED,
923
InventoryEntry.MODIFIED_AND_RENAMED):
924
old_path = self.basis_inv.id2path(ie.file_id)
925
self.reporter.renamed(change, old_path, path)
926
self._next_progress_entry()
928
if change == 'unchanged':
930
self.reporter.snapshot_change(change, path)
931
self._next_progress_entry()
933
def _set_progress_stage(self, name, counter=False):
934
"""Set the progress stage and emit an update to the progress bar."""
935
self.pb_stage_name = name
936
self.pb_stage_count += 1
938
self.pb_entries_count = 0
940
self.pb_entries_count = None
941
self._emit_progress()
943
def _next_progress_entry(self):
944
"""Emit an update to the progress bar and increment the entry count."""
945
self.pb_entries_count += 1
946
self._emit_progress()
948
def _emit_progress(self):
949
if self.pb_entries_count is not None:
950
text = "%s [%d] - Stage" % (self.pb_stage_name,
951
self.pb_entries_count)
953
text = "%s - Stage" % (self.pb_stage_name, )
954
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
956
def _set_specific_file_ids(self):
957
"""populate self.specific_file_ids if we will use it."""
958
if not self.use_record_iter_changes:
959
# If provided, ensure the specified files are versioned
960
if self.specific_files is not None:
961
# Note: This routine is being called because it raises
962
# PathNotVersionedError as a side effect of finding the IDs. We
963
# later use the ids we found as input to the working tree
964
# inventory iterator, so we only consider those ids rather than
965
# examining the whole tree again.
966
# XXX: Dont we have filter_unversioned to do this more
968
self.specific_file_ids = tree.find_ids_across_trees(
969
self.specific_files, [self.basis_tree, self.work_tree])
460
note('modified %s', path)
461
self._commit_file(new_ie, file_id, file_parents)
464
def _commit_nonfile(self, file_id):
465
self.new_inv.add(self.work_inv[file_id].copy())
468
def _carry_file(self, file_id):
469
"""Carry the file unchanged from the basis revision."""
470
if self.basis_inv.has_id(file_id):
471
self.new_inv.add(self.basis_inv[file_id].copy())
474
def _commit_old_text(self, file_id, parent_ie):
475
"""Keep the same text as last time, but possibly a different name."""
476
ie = self.work_inv[file_id].copy()
477
ie.text_version = parent_ie.text_version
478
ie.text_size = parent_ie.text_size
479
ie.text_sha1 = parent_ie.text_sha1
483
def _report_deletes(self):
484
for file_id in self.basis_inv:
485
if file_id not in self.new_inv:
486
note('deleted %s', self.basis_inv.id2path(file_id))
489
def _commit_file(self, new_ie, file_id, file_parents):
490
mutter('store new text for {%s} in revision {%s}',
491
file_id, self.rev_id)
492
new_lines = self.work_tree.get_file(file_id).readlines()
493
self._add_text_to_weave(file_id, new_lines, file_parents)
494
new_ie.text_version = self.rev_id
495
new_ie.text_sha1 = sha_strings(new_lines)
496
new_ie.text_size = sum(map(len, new_lines))
497
self.new_inv.add(new_ie)
500
def _add_text_to_weave(self, file_id, new_lines, parents):
501
self.weave_store.add_text(file_id, self.rev_id, new_lines, parents)
504
def _gen_revision_id(branch, when):
505
"""Return new revision-id."""
506
s = '%s-%s-' % (user_email(branch), compact_date(when))
507
s += hexlify(rand_bytes(8))
513
def merge_ancestry_lines(rev_id, ancestries):
514
"""Return merged ancestry lines.
516
rev_id -- id of the new revision
518
ancestries -- a sequence of ancestries for parent revisions,
519
as newline-terminated line lists.
521
if len(ancestries) == 0:
522
return [rev_id + '\n']
523
seen = set(ancestries[0])
524
ancs = ancestries[0][:]
525
for parent_ancestry in ancestries[1:]:
526
for line in parent_ancestry:
527
assert line[-1] == '\n'
971
self.specific_file_ids = None
added
removed
bzrlib/commit.py
