1
# Copyright (C) 2005-2011 Canonical Ltd
1
# Copyright (C) 2005 Canonical Ltd
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
5
5
# the Free Software Foundation; either version 2 of the License, or
6
6
# (at your option) any later version.
8
8
# This program is distributed in the hope that it will be useful,
9
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
11
# GNU General Public License for more details.
13
13
# You should have received a copy of the GNU General Public License
14
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
19
# The newly committed revision is going to have a shape corresponding
20
# to that of the working tree. Files that are not in the
21
# working tree and that were in the predecessor are reported as
22
# removed --- this can include files that were either removed from the
23
# inventory or deleted in the working tree. If they were only
24
# deleted from disk, they are removed from the working inventory.
26
# We then consider the remaining entries, which will be in the new
27
# version. Directory entries are simply copied across. File entries
28
# must be checked to see if a new version of the file should be
29
# recorded. For each parent revision tree, we check to see what
30
# version of the file was present. If the file was present in at
31
# least one tree, and if it was the same version in all the trees,
32
# then we can just refer to that version. Otherwise, a new version
33
# representing the merger of the file versions must be added.
35
# TODO: Update hashcache before and after - or does the WorkingTree
38
# TODO: Rather than mashing together the ancestry and storing it back,
39
# perhaps the weave should have single method which does it all in one
40
# go, avoiding a lot of redundant work.
42
# TODO: Perhaps give a warning if one of the revisions marked as
43
# merged is already in the ancestry, and then don't record it as a
46
# TODO: If the file is newly merged but unchanged from the version it
47
# merges from, then it should still be reported as newly added
48
# relative to the basis revision.
50
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
51
# the rest of the code; add a deprecation of the old name.
60
from bzrlib.branch import Branch
61
from bzrlib.cleanup import OperationWithCleanups
63
from bzrlib.errors import (BzrError, PointlessCommit,
67
from bzrlib.osutils import (get_user_encoding,
69
minimum_path_selection,
72
from bzrlib.trace import mutter, note, is_quiet
73
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
74
from bzrlib import symbol_versioning
75
from bzrlib.urlutils import unescape_for_display
76
from bzrlib.i18n import gettext
78
class NullCommitReporter(object):
79
"""I report on progress of a commit."""
81
def started(self, revno, revid, location=None):
83
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
84
"to started.", DeprecationWarning,
88
def snapshot_change(self, change, path):
91
def completed(self, revno, rev_id):
94
def deleted(self, path):
97
def missing(self, path):
100
def renamed(self, change, old_path, new_path):
103
def is_verbose(self):
107
class ReportCommitToLog(NullCommitReporter):
109
def _note(self, format, *args):
112
Subclasses may choose to override this method.
116
def snapshot_change(self, change, path):
117
if path == '' and change in (gettext('added'), gettext('modified')):
119
self._note("%s %s", change, path)
121
def started(self, revno, rev_id, location=None):
122
if location is not None:
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,
131
self._note(gettext('Committing%s'), location)
133
def completed(self, revno, rev_id):
134
self._note(gettext('Committed revision %d.'), revno)
135
# self._note goes to the console too; so while we want to log the
136
# rev_id, we can't trivially only log it. (See bug 526425). Long
137
# term we should rearrange the reporting structure, but for now
138
# we just mutter seperately. We mutter the revid and revno together
139
# so that concurrent bzr invocations won't lead to confusion.
140
mutter('Committed revid %s as revno %d.', rev_id, revno)
142
def deleted(self, path):
143
self._note(gettext('deleted %s'), path)
145
def missing(self, path):
146
self._note(gettext('missing %s'), path)
148
def renamed(self, change, old_path, new_path):
149
self._note('%s %s => %s', change, old_path, new_path)
151
def is_verbose(self):
155
class Commit(object):
156
"""Task of committing a new revision.
158
This is a MethodObject: it accumulates state as the commit is
159
prepared, and then it is discarded. It doesn't represent
160
historical revisions, just the act of recording a new one.
163
Modified to hold a list of files that have been deleted from
164
the working directory; these should be removed from the
170
"""Create a Commit object.
172
:param reporter: the default reporter to use or None to decide later
174
self.reporter = reporter
175
self.config_stack = config_stack
178
def update_revprops(revprops, branch, authors=None, author=None,
179
local=False, possible_master_transports=None):
182
if possible_master_transports is None:
183
possible_master_transports = []
184
if not 'branch-nick' in revprops:
185
revprops['branch-nick'] = branch._get_nick(
187
possible_master_transports)
188
if authors is not None:
189
if author is not None:
190
raise AssertionError('Specifying both author and authors '
191
'is not allowed. Specify just authors instead')
192
if 'author' in revprops or 'authors' in revprops:
193
# XXX: maybe we should just accept one of them?
194
raise AssertionError('author property given twice')
196
for individual in authors:
197
if '\n' in individual:
198
raise AssertionError('\\n is not a valid character '
199
'in an author identity')
200
revprops['authors'] = '\n'.join(authors)
201
if author is not None:
202
symbol_versioning.warn('The parameter author was deprecated'
203
' in version 1.13. Use authors instead',
205
if 'author' in revprops or 'authors' in revprops:
206
# XXX: maybe we should just accept one of them?
207
raise AssertionError('author property given twice')
209
raise AssertionError('\\n is not a valid character '
210
'in an author identity')
211
revprops['authors'] = author
221
allow_pointless=True,
229
message_callback=None,
232
possible_master_transports=None,
234
"""Commit working copy as a new revision.
236
:param message: the commit message (it or message_callback is required)
237
:param message_callback: A callback: message = message_callback(cmt_obj)
239
:param timestamp: if not None, seconds-since-epoch for a
240
postdated/predated commit.
242
:param specific_files: If not None, commit only those files. An empty
243
list means 'commit no files'.
245
:param rev_id: If set, use this as the new revision id.
246
Useful for test or import commands that need to tightly
247
control what revisions are assigned. If you duplicate
248
a revision id that exists elsewhere it is your own fault.
249
If null (default), a time/random revision id is generated.
251
:param allow_pointless: If true (default), commit even if nothing
252
has changed and no merges are recorded.
254
:param strict: If true, don't allow a commit if the working tree
255
contains unknown files.
257
:param revprops: Properties for new revision
258
:param local: Perform a local only commit.
259
:param reporter: the reporter to use or None for the default
260
:param verbose: if True and the reporter is not None, report everything
261
:param recursive: If set to 'down', commit in any subtrees that have
262
pending changes of any sort during this commit.
263
:param exclude: None or a list of relative paths to exclude from the
264
commit. Pending changes to excluded files will be ignored by the
266
:param lossy: When committing to a foreign VCS, ignore any
267
data that can not be natively represented.
269
operation = OperationWithCleanups(self._commit)
270
self.revprops = revprops or {}
271
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
272
self.config_stack = config or self.config_stack
273
return operation.run(
278
specific_files=specific_files,
280
allow_pointless=allow_pointless,
283
working_tree=working_tree,
286
message_callback=message_callback,
289
possible_master_transports=possible_master_transports,
292
def _commit(self, operation, message, timestamp, timezone, committer,
293
specific_files, rev_id, allow_pointless, strict, verbose,
294
working_tree, local, reporter, message_callback, recursive,
295
exclude, possible_master_transports, lossy):
296
mutter('preparing to commit')
298
if working_tree is None:
299
raise BzrError("working_tree must be passed into commit().")
301
self.work_tree = working_tree
302
self.branch = self.work_tree.branch
303
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
304
if not self.branch.repository.supports_rich_root():
305
raise errors.RootNotRich()
306
if message_callback is None:
307
if message is not None:
308
if isinstance(message, str):
309
message = message.decode(get_user_encoding())
310
message_callback = lambda x: message
312
raise BzrError("The message or message_callback keyword"
313
" parameter is required for commit().")
315
self.bound_branch = None
316
self.any_entries_deleted = False
317
if exclude is not None:
318
self.exclude = sorted(
319
minimum_path_selection(exclude))
323
self.master_branch = None
324
self.recursive = recursive
326
# self.specific_files is None to indicate no filter, or any iterable to
327
# indicate a filter - [] means no files at all, as per iter_changes.
328
if specific_files is not None:
329
self.specific_files = sorted(
330
minimum_path_selection(specific_files))
332
self.specific_files = None
334
self.allow_pointless = allow_pointless
335
self.message_callback = message_callback
336
self.timestamp = timestamp
337
self.timezone = timezone
338
self.committer = committer
340
self.verbose = verbose
342
self.work_tree.lock_write()
343
operation.add_cleanup(self.work_tree.unlock)
344
self.parents = self.work_tree.get_parent_ids()
345
# We can use record_iter_changes IFF iter_changes is compatible with
346
# the command line parameters, and the repository has fast delta
347
# generation. See bug 347649.
348
self.use_record_iter_changes = (
350
not self.branch.repository._format.supports_tree_reference and
351
(self.branch.repository._format.fast_deltas or
352
len(self.parents) < 2))
353
self.pb = ui.ui_factory.nested_progress_bar()
354
operation.add_cleanup(self.pb.finished)
355
self.basis_revid = self.work_tree.last_revision()
356
self.basis_tree = self.work_tree.basis_tree()
357
self.basis_tree.lock_read()
358
operation.add_cleanup(self.basis_tree.unlock)
359
# Cannot commit with conflicts present.
360
if len(self.work_tree.conflicts()) > 0:
361
raise ConflictsInTree
363
# Setup the bound branch variables as needed.
364
self._check_bound_branch(operation, possible_master_transports)
366
# Check that the working tree is up to date
367
old_revno, old_revid, new_revno = self._check_out_of_date_tree()
369
# Complete configuration setup
370
if reporter is not None:
371
self.reporter = reporter
372
elif self.reporter is None:
373
self.reporter = self._select_reporter()
374
if self.config_stack is None:
375
self.config_stack = self.work_tree.get_config_stack()
377
self._set_specific_file_ids()
379
# Setup the progress bar. As the number of files that need to be
380
# committed in unknown, progress is reported as stages.
381
# We keep track of entries separately though and include that
382
# information in the progress bar during the relevant stages.
383
self.pb_stage_name = ""
384
self.pb_stage_count = 0
385
self.pb_stage_total = 5
386
if self.bound_branch:
387
# 2 extra stages: "Uploading data to master branch" and "Merging
388
# tags to master branch"
389
self.pb_stage_total += 2
390
self.pb.show_pct = False
391
self.pb.show_spinner = False
392
self.pb.show_eta = False
393
self.pb.show_count = True
394
self.pb.show_bar = True
396
self._gather_parents()
397
# After a merge, a selected file commit is not supported.
398
# See 'bzr help merge' for an explanation as to why.
399
if len(self.parents) > 1 and self.specific_files is not None:
400
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
401
# Excludes are a form of selected file commit.
402
if len(self.parents) > 1 and self.exclude:
403
raise errors.CannotCommitSelectedFileMerge(self.exclude)
405
# Collect the changes
406
self._set_progress_stage("Collecting changes", counter=True)
408
self.builder = self.branch.get_commit_builder(self.parents,
409
self.config_stack, timestamp, timezone, committer, self.revprops,
411
if not self.builder.supports_record_entry_contents and self.exclude:
413
raise errors.ExcludesUnsupported(self.branch.repository)
415
if self.builder.updates_branch and self.bound_branch:
417
raise AssertionError(
418
"bound branches not supported for commit builders "
419
"that update the branch")
422
self.builder.will_record_deletes()
423
# find the location being committed to
424
if self.bound_branch:
425
master_location = self.master_branch.base
427
master_location = self.branch.base
429
# report the start of the commit
430
self.reporter.started(new_revno, self.rev_id, master_location)
432
self._update_builder_with_changes()
433
self._check_pointless()
435
# TODO: Now the new inventory is known, check for conflicts.
436
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
437
# weave lines, because nothing should be recorded until it is known
438
# that commit will succeed.
439
self._set_progress_stage("Saving data locally")
440
self.builder.finish_inventory()
442
# Prompt the user for a commit message if none provided
443
message = message_callback(self)
444
self.message = message
446
# Add revision data to the local branch
447
self.rev_id = self.builder.commit(self.message)
450
mutter("aborting commit write group because of exception:")
451
trace.log_exception_quietly()
455
self._update_branches(old_revno, old_revid, new_revno)
457
# Make the working tree be up to date with the branch. This
458
# includes automatic changes scheduled to be made to the tree, such
459
# as updating its basis and unversioning paths that were missing.
460
self.work_tree.unversion(self.deleted_ids)
461
self._set_progress_stage("Updating the working tree")
462
self.work_tree.update_basis_by_delta(self.rev_id,
463
self.builder.get_basis_delta())
464
self.reporter.completed(new_revno, self.rev_id)
465
self._process_post_hooks(old_revno, new_revno)
468
def _update_branches(self, old_revno, old_revid, new_revno):
469
"""Update the master and local branch to the new revision.
471
This will try to make sure that the master branch is updated
472
before the local branch.
474
:param old_revno: Revision number of master branch before the
476
:param old_revid: Tip of master branch before the commit
477
:param new_revno: Revision number of the new commit
479
if not self.builder.updates_branch:
480
self._process_pre_hooks(old_revno, new_revno)
482
# Upload revision data to the master.
483
# this will propagate merged revisions too if needed.
484
if self.bound_branch:
485
self._set_progress_stage("Uploading data to master branch")
486
# 'commit' to the master first so a timeout here causes the
487
# local branch to be out of date
488
(new_revno, self.rev_id) = self.master_branch.import_last_revision_info_and_tags(
489
self.branch, new_revno, self.rev_id, lossy=self._lossy)
491
self.branch.fetch(self.master_branch, self.rev_id)
493
# and now do the commit locally.
494
self.branch.set_last_revision_info(new_revno, self.rev_id)
497
self._process_pre_hooks(old_revno, new_revno)
499
# The commit builder will already have updated the branch,
501
self.branch.set_last_revision_info(old_revno, old_revid)
504
# Merge local tags to remote
505
if self.bound_branch:
506
self._set_progress_stage("Merging tags to master branch")
507
tag_updates, tag_conflicts = self.branch.tags.merge_to(
508
self.master_branch.tags)
510
warning_lines = [' ' + name for name, _, _ in tag_conflicts]
511
note( gettext("Conflicting tags in bound branch:\n{0}".format(
512
"\n".join(warning_lines))) )
514
def _select_reporter(self):
515
"""Select the CommitReporter to use."""
517
return NullCommitReporter()
518
return ReportCommitToLog()
520
def _check_pointless(self):
521
if self.allow_pointless:
523
# A merge with no effect on files
524
if len(self.parents) > 1:
526
if self.builder.any_changes():
528
raise PointlessCommit()
530
def _check_bound_branch(self, operation, possible_master_transports=None):
531
"""Check to see if the local branch is bound.
533
If it is bound, then most of the commit will actually be
534
done using the remote branch as the target branch.
535
Only at the end will the local branch be updated.
537
if self.local and not self.branch.get_bound_location():
538
raise errors.LocalRequiresBoundBranch()
541
self.master_branch = self.branch.get_master_branch(
542
possible_master_transports)
544
if not self.master_branch:
545
# make this branch the reference branch for out of date checks.
546
self.master_branch = self.branch
549
# If the master branch is bound, we must fail
550
master_bound_location = self.master_branch.get_bound_location()
551
if master_bound_location:
552
raise errors.CommitToDoubleBoundBranch(self.branch,
553
self.master_branch, master_bound_location)
555
# TODO: jam 20051230 We could automatically push local
556
# commits to the remote branch if they would fit.
557
# But for now, just require remote to be identical
560
# Make sure the local branch is identical to the master
561
master_info = self.master_branch.last_revision_info()
562
local_info = self.branch.last_revision_info()
563
if local_info != master_info:
564
raise errors.BoundBranchOutOfDate(self.branch,
567
# Now things are ready to change the master branch
569
self.bound_branch = self.branch
570
self.master_branch.lock_write()
571
operation.add_cleanup(self.master_branch.unlock)
573
def _check_out_of_date_tree(self):
574
"""Check that the working tree is up to date.
576
:return: old_revision_number, old_revision_id, new_revision_number
580
first_tree_parent = self.work_tree.get_parent_ids()[0]
582
# if there are no parents, treat our parent as 'None'
583
# this is so that we still consider the master branch
584
# - in a checkout scenario the tree may have no
585
# parents but the branch may do.
586
first_tree_parent = bzrlib.revision.NULL_REVISION
587
old_revno, master_last = self.master_branch.last_revision_info()
588
if master_last != first_tree_parent:
589
if master_last != bzrlib.revision.NULL_REVISION:
590
raise errors.OutOfDateTree(self.work_tree)
591
if self.branch.repository.has_revision(first_tree_parent):
592
new_revno = old_revno + 1
594
# ghost parents never appear in revision history.
596
return old_revno, master_last, new_revno
598
def _process_pre_hooks(self, old_revno, new_revno):
599
"""Process any registered pre commit hooks."""
600
self._set_progress_stage("Running pre_commit hooks")
601
self._process_hooks("pre_commit", old_revno, new_revno)
603
def _process_post_hooks(self, old_revno, new_revno):
604
"""Process any registered post commit hooks."""
605
# Process the post commit hooks, if any
606
self._set_progress_stage("Running post_commit hooks")
607
# old style commit hooks - should be deprecated ? (obsoleted in
608
# 0.15^H^H^H^H 2.5.0)
609
post_commit = self.config_stack.get('post_commit')
610
if post_commit is not None:
611
hooks = post_commit.split(' ')
612
# this would be nicer with twisted.python.reflect.namedAny
614
result = eval(hook + '(branch, rev_id)',
615
{'branch':self.branch,
617
'rev_id':self.rev_id})
618
# process new style post commit hooks
619
self._process_hooks("post_commit", old_revno, new_revno)
621
def _process_hooks(self, hook_name, old_revno, new_revno):
622
if not Branch.hooks[hook_name]:
625
# new style commit hooks:
626
if not self.bound_branch:
627
hook_master = self.branch
630
hook_master = self.master_branch
631
hook_local = self.branch
632
# With bound branches, when the master is behind the local branch,
633
# the 'old_revno' and old_revid values here are incorrect.
634
# XXX: FIXME ^. RBC 20060206
636
old_revid = self.parents[0]
638
old_revid = bzrlib.revision.NULL_REVISION
640
if hook_name == "pre_commit":
641
future_tree = self.builder.revision_tree()
642
tree_delta = future_tree.changes_from(self.basis_tree,
645
for hook in Branch.hooks[hook_name]:
646
# show the running hook in the progress bar. As hooks may
647
# end up doing nothing (e.g. because they are not configured by
648
# the user) this is still showing progress, not showing overall
649
# actions - its up to each plugin to show a UI if it want's to
650
# (such as 'Emailing diff to foo@example.com').
651
self.pb_stage_name = "Running %s hooks [%s]" % \
652
(hook_name, Branch.hooks.get_hook_name(hook))
653
self._emit_progress()
654
if 'hooks' in debug.debug_flags:
655
mutter("Invoking commit hook: %r", hook)
656
if hook_name == "post_commit":
657
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
659
elif hook_name == "pre_commit":
660
hook(hook_local, hook_master,
661
old_revno, old_revid, new_revno, self.rev_id,
662
tree_delta, future_tree)
664
def _gather_parents(self):
665
"""Record the parents of a merge for merge detection."""
666
# TODO: Make sure that this list doesn't contain duplicate
667
# entries and the order is preserved when doing this.
668
if self.use_record_iter_changes:
670
self.basis_inv = self.basis_tree.root_inventory
671
self.parent_invs = [self.basis_inv]
672
for revision in self.parents[1:]:
673
if self.branch.repository.has_revision(revision):
674
mutter('commit parent revision {%s}', revision)
675
inventory = self.branch.repository.get_inventory(revision)
676
self.parent_invs.append(inventory)
678
mutter('commit parent ghost revision {%s}', revision)
680
def _update_builder_with_changes(self):
681
"""Update the commit builder with the data about what has changed.
683
exclude = self.exclude
684
specific_files = self.specific_files
685
mutter("Selecting files for commit with filter %s", specific_files)
688
if self.use_record_iter_changes:
689
iter_changes = self.work_tree.iter_changes(self.basis_tree,
690
specific_files=specific_files)
691
iter_changes = self._filter_iter_changes(iter_changes)
692
for file_id, path, fs_hash in self.builder.record_iter_changes(
693
self.work_tree, self.basis_revid, iter_changes):
694
self.work_tree._observed_sha1(file_id, path, fs_hash)
696
# Build the new inventory
697
self._populate_from_inventory()
698
self._record_unselected()
699
self._report_and_accumulate_deletes()
701
def _filter_iter_changes(self, iter_changes):
702
"""Process iter_changes.
704
This method reports on the changes in iter_changes to the user, and
705
converts 'missing' entries in the iter_changes iterator to 'deleted'
706
entries. 'missing' entries have their
708
:param iter_changes: An iter_changes to process.
709
:return: A generator of changes.
711
reporter = self.reporter
712
report_changes = reporter.is_verbose()
714
for change in iter_changes:
716
old_path = change[1][0]
717
new_path = change[1][1]
718
versioned = change[3][1]
720
versioned = change[3][1]
721
if kind is None and versioned:
724
reporter.missing(new_path)
725
deleted_ids.append(change[0])
726
# Reset the new path (None) and new versioned flag (False)
727
change = (change[0], (change[1][0], None), change[2],
728
(change[3][0], False)) + change[4:]
729
new_path = change[1][1]
731
elif kind == 'tree-reference':
732
if self.recursive == 'down':
733
self._commit_nested_tree(change[0], change[1][1])
734
if change[3][0] or change[3][1]:
738
reporter.deleted(old_path)
739
elif old_path is None:
740
reporter.snapshot_change(gettext('added'), new_path)
741
elif old_path != new_path:
742
reporter.renamed(gettext('renamed'), old_path, new_path)
745
self.work_tree.branch.repository._format.rich_root_data):
746
# Don't report on changes to '' in non rich root
748
reporter.snapshot_change(gettext('modified'), new_path)
749
self._next_progress_entry()
750
# Unversion IDs that were found to be deleted
751
self.deleted_ids = deleted_ids
753
def _record_unselected(self):
754
# If specific files are selected, then all un-selected files must be
755
# recorded in their previous state. For more details, see
756
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
757
if self.specific_files or self.exclude:
758
specific_files = self.specific_files or []
759
for path, old_ie in self.basis_inv.iter_entries():
760
if self.builder.new_inventory.has_id(old_ie.file_id):
761
# already added - skip.
763
if (is_inside_any(specific_files, path)
764
and not is_inside_any(self.exclude, path)):
765
# was inside the selected path, and not excluded - if not
766
# present it has been deleted so skip.
768
# From here down it was either not selected, or was excluded:
769
# We preserve the entry unaltered.
771
# Note: specific file commits after a merge are currently
772
# prohibited. This test is for sanity/safety in case it's
773
# required after that changes.
774
if len(self.parents) > 1:
776
self.builder.record_entry_contents(ie, self.parent_invs, path,
777
self.basis_tree, None)
779
def _report_and_accumulate_deletes(self):
780
if (isinstance(self.basis_inv, Inventory)
781
and isinstance(self.builder.new_inventory, Inventory)):
782
# the older Inventory classes provide a _byid dict, and building a
783
# set from the keys of this dict is substantially faster than even
784
# getting a set of ids from the inventory
786
# <lifeless> set(dict) is roughly the same speed as
787
# set(iter(dict)) and both are significantly slower than
789
deleted_ids = set(self.basis_inv._byid.keys()) - \
790
set(self.builder.new_inventory._byid.keys())
792
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
794
self.any_entries_deleted = True
795
deleted = [(self.basis_tree.id2path(file_id), file_id)
796
for file_id in deleted_ids]
798
# XXX: this is not quite directory-order sorting
799
for path, file_id in deleted:
800
self.builder.record_delete(path, file_id)
801
self.reporter.deleted(path)
803
def _check_strict(self):
804
# XXX: when we use iter_changes this would likely be faster if
805
# iter_changes would check for us (even in the presence of
808
# raise an exception as soon as we find a single unknown.
809
for unknown in self.work_tree.unknowns():
810
raise StrictCommitFailed()
812
def _populate_from_inventory(self):
813
"""Populate the CommitBuilder by walking the working tree inventory."""
814
# Build the revision inventory.
816
# This starts by creating a new empty inventory. Depending on
817
# which files are selected for commit, and what is present in the
818
# current tree, the new inventory is populated. inventory entries
819
# which are candidates for modification have their revision set to
820
# None; inventory entries that are carried over untouched have their
821
# revision set to their prior value.
823
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
824
# results to create a new inventory at the same time, which results
825
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
828
specific_files = self.specific_files
829
exclude = self.exclude
830
report_changes = self.reporter.is_verbose()
832
# A tree of paths that have been deleted. E.g. if foo/bar has been
833
# deleted, then we have {'foo':{'bar':{}}}
835
# XXX: Note that entries may have the wrong kind because the entry does
836
# not reflect the status on disk.
837
# NB: entries will include entries within the excluded ids/paths
838
# because iter_entries_by_dir has no 'exclude' facility today.
839
entries = self.work_tree.iter_entries_by_dir(
840
specific_file_ids=self.specific_file_ids, yield_parents=True)
841
for path, existing_ie in entries:
842
file_id = existing_ie.file_id
843
name = existing_ie.name
844
parent_id = existing_ie.parent_id
845
kind = existing_ie.kind
846
# Skip files that have been deleted from the working tree.
847
# The deleted path ids are also recorded so they can be explicitly
850
path_segments = splitpath(path)
851
deleted_dict = deleted_paths
852
for segment in path_segments:
853
deleted_dict = deleted_dict.get(segment, None)
855
# We either took a path not present in the dict
856
# (deleted_dict was None), or we've reached an empty
857
# child dir in the dict, so are now a sub-path.
861
if deleted_dict is not None:
862
# the path has a deleted parent, do not add it.
864
if exclude and is_inside_any(exclude, path):
865
# Skip excluded paths. Excluded paths are processed by
866
# _update_builder_with_changes.
868
content_summary = self.work_tree.path_content_summary(path)
869
kind = content_summary[0]
870
# Note that when a filter of specific files is given, we must only
871
# skip/record deleted files matching that filter.
872
if not specific_files or is_inside_any(specific_files, path):
873
if kind == 'missing':
874
if not deleted_paths:
875
# path won't have been split yet.
876
path_segments = splitpath(path)
877
deleted_dict = deleted_paths
878
for segment in path_segments:
879
deleted_dict = deleted_dict.setdefault(segment, {})
880
self.reporter.missing(path)
881
self._next_progress_entry()
882
deleted_ids.append(file_id)
884
# TODO: have the builder do the nested commit just-in-time IF and
886
if kind == 'tree-reference':
887
# enforce repository nested tree policy.
888
if (not self.work_tree.supports_tree_reference() or
889
# repository does not support it either.
890
not self.branch.repository._format.supports_tree_reference):
892
content_summary = (kind, None, None, None)
893
elif self.recursive == 'down':
894
nested_revision_id = self._commit_nested_tree(
896
content_summary = (kind, None, None, nested_revision_id)
898
nested_revision_id = self.work_tree.get_reference_revision(file_id)
899
content_summary = (kind, None, None, nested_revision_id)
901
# Record an entry for this item
902
# Note: I don't particularly want to have the existing_ie
903
# parameter but the test suite currently (28-Jun-07) breaks
904
# without it thanks to a unicode normalisation issue. :-(
905
definitely_changed = kind != existing_ie.kind
906
self._record_entry(path, file_id, specific_files, kind, name,
907
parent_id, definitely_changed, existing_ie, report_changes,
910
# Unversion IDs that were found to be deleted
911
self.deleted_ids = deleted_ids
913
def _commit_nested_tree(self, file_id, path):
914
"Commit a nested tree."
915
sub_tree = self.work_tree.get_nested_tree(file_id, path)
916
# FIXME: be more comprehensive here:
917
# this works when both trees are in --trees repository,
918
# but when both are bound to a different repository,
919
# it fails; a better way of approaching this is to
920
# finally implement the explicit-caches approach design
921
# a while back - RBC 20070306.
922
if sub_tree.branch.repository.has_same_location(
923
self.work_tree.branch.repository):
924
sub_tree.branch.repository = \
925
self.work_tree.branch.repository
927
return sub_tree.commit(message=None, revprops=self.revprops,
928
recursive=self.recursive,
929
message_callback=self.message_callback,
930
timestamp=self.timestamp, timezone=self.timezone,
931
committer=self.committer,
932
allow_pointless=self.allow_pointless,
933
strict=self.strict, verbose=self.verbose,
934
local=self.local, reporter=self.reporter)
935
except errors.PointlessCommit:
936
return self.work_tree.get_reference_revision(file_id)
938
def _record_entry(self, path, file_id, specific_files, kind, name,
939
parent_id, definitely_changed, existing_ie, report_changes,
941
"Record the new inventory entry for a path if any."
942
# mutter('check %s {%s}', path, file_id)
943
# mutter('%s selected for commit', path)
944
if definitely_changed or existing_ie is None:
945
ie = make_entry(kind, name, parent_id, file_id)
947
ie = existing_ie.copy()
949
# For carried over entries we don't care about the fs hash - the repo
950
# isn't generating a sha, so we're not saving computation time.
951
_, _, fs_hash = self.builder.record_entry_contents(
952
ie, self.parent_invs, path, self.work_tree, content_summary)
954
self._report_change(ie, path)
956
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
959
def _report_change(self, ie, path):
960
"""Report a change to the user.
962
The change that has occurred is described relative to the basis
965
if (self.basis_inv.has_id(ie.file_id)):
966
basis_ie = self.basis_inv[ie.file_id]
969
change = ie.describe_change(basis_ie, ie)
970
if change in (InventoryEntry.RENAMED,
971
InventoryEntry.MODIFIED_AND_RENAMED):
972
old_path = self.basis_inv.id2path(ie.file_id)
973
self.reporter.renamed(change, old_path, path)
974
self._next_progress_entry()
976
if change == gettext('unchanged'):
978
self.reporter.snapshot_change(change, path)
979
self._next_progress_entry()
981
def _set_progress_stage(self, name, counter=False):
982
"""Set the progress stage and emit an update to the progress bar."""
983
self.pb_stage_name = name
984
self.pb_stage_count += 1
986
self.pb_entries_count = 0
988
self.pb_entries_count = None
989
self._emit_progress()
991
def _next_progress_entry(self):
992
"""Emit an update to the progress bar and increment the entry count."""
993
self.pb_entries_count += 1
994
self._emit_progress()
996
def _emit_progress(self):
997
if self.pb_entries_count is not None:
998
text = gettext("{0} [{1}] - Stage").format(self.pb_stage_name,
999
self.pb_entries_count)
1001
text = gettext("%s - Stage") % (self.pb_stage_name, )
1002
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
1004
def _set_specific_file_ids(self):
1005
"""populate self.specific_file_ids if we will use it."""
1006
if not self.use_record_iter_changes:
1007
# If provided, ensure the specified files are versioned
1008
if self.specific_files is not None:
1009
# Note: This routine is being called because it raises
1010
# PathNotVersionedError as a side effect of finding the IDs. We
1011
# later use the ids we found as input to the working tree
1012
# inventory iterator, so we only consider those ids rather than
1013
# examining the whole tree again.
1014
# XXX: Dont we have filter_unversioned to do this more
1016
self.specific_file_ids = tree.find_ids_across_trees(
1017
self.specific_files, [self.basis_tree, self.work_tree])
1019
self.specific_file_ids = None
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
def commit(branch, message,
26
allow_pointless=True):
27
"""Commit working copy as a new revision.
29
The basic approach is to add all the file texts into the
30
store, then the inventory, then make a new revision pointing
31
to that inventory and store that.
33
This is not quite safe if the working copy changes during the
34
commit; for the moment that is simply not allowed. A better
35
approach is to make a temporary copy of the files before
36
computing their hashes, and then add those hashes in turn to
37
the inventory. This should mean at least that there are no
38
broken hash pointers. There is no way we can get a snapshot
39
of the whole directory at an instant. This would also have to
40
be robust against files disappearing, moving, etc. So the
41
whole thing is a bit hard.
43
This raises PointlessCommit if there are no changes, no new merges,
44
and allow_pointless is false.
46
timestamp -- if not None, seconds-since-epoch for a
47
postdated/predated commit.
50
If true, commit only those files.
53
If set, use this as the new revision id.
54
Useful for test or import commands that need to tightly
55
control what revisions are assigned. If you duplicate
56
a revision id that exists elsewhere it is your own fault.
57
If null (default), a time/random revision id is generated.
62
from bzrlib.osutils import local_time_offset, username
63
from bzrlib.branch import gen_file_id
64
from bzrlib.errors import BzrError, PointlessCommit
65
from bzrlib.revision import Revision, RevisionReference
66
from bzrlib.trace import mutter, note
67
from bzrlib.xml import serializer_v4
72
# First walk over the working inventory; and both update that
73
# and also build a new revision inventory. The revision
74
# inventory needs to hold the text-id, sha1 and size of the
75
# actual file versions committed in the revision. (These are
76
# not present in the working inventory.) We also need to
77
# detect missing/deleted files, and remove them from the
80
work_tree = branch.working_tree()
81
work_inv = work_tree.inventory
82
basis = branch.basis_tree()
83
basis_inv = basis.inventory
86
# note('looking for changes...')
87
# print 'looking for changes...'
88
# disabled; should be done at a higher level
91
pending_merges = branch.pending_merges()
93
missing_ids, new_inv, any_changes = \
94
_gather_commit(branch,
101
if not (any_changes or allow_pointless or pending_merges):
102
raise PointlessCommit()
104
for file_id in missing_ids:
105
# Any files that have been deleted are now removed from the
106
# working inventory. Files that were not selected for commit
107
# are left as they were in the working inventory and ommitted
108
# from the revision inventory.
110
# have to do this later so we don't mess up the iterator.
111
# since parents may be removed before their children we
114
# FIXME: There's probably a better way to do this; perhaps
115
# the workingtree should know how to filter itbranch.
116
if work_inv.has_id(file_id):
117
del work_inv[file_id]
120
rev_id = _gen_revision_id(branch, time.time())
123
inv_tmp = tempfile.TemporaryFile()
125
serializer_v4.write_inventory(new_inv, inv_tmp)
127
branch.inventory_store.add(inv_tmp, inv_id)
128
mutter('new inventory_id is {%s}' % inv_id)
130
# We could also just sha hash the inv_tmp file
131
# however, in the case that branch.inventory_store.add()
132
# ever actually does anything special
133
inv_sha1 = branch.get_inventory_sha1(inv_id)
135
branch._write_inventory(work_inv)
137
if timestamp == None:
138
timestamp = time.time()
140
if committer == None:
141
committer = username(branch)
144
timezone = local_time_offset()
146
mutter("building commit log message")
147
rev = Revision(timestamp=timestamp,
152
inventory_sha1=inv_sha1,
156
precursor_id = branch.last_patch()
158
precursor_sha1 = branch.get_revision_sha1(precursor_id)
159
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
160
for merge_rev in pending_merges:
161
rev.parents.append(RevisionReference(merge_rev))
163
rev_tmp = tempfile.TemporaryFile()
164
serializer_v4.write_revision(rev, rev_tmp)
166
branch.revision_store.add(rev_tmp, rev_id)
167
mutter("new revision_id is {%s}" % rev_id)
169
## XXX: Everything up to here can simply be orphaned if we abort
170
## the commit; it will leave junk files behind but that doesn't
173
## TODO: Read back the just-generated changeset, and make sure it
174
## applies and recreates the right state.
176
## TODO: Also calculate and store the inventory SHA1
177
mutter("committing patch r%d" % (branch.revno() + 1))
179
branch.append_revision(rev_id)
181
branch.set_pending_merges([])
184
# disabled; should go through logging
185
# note("commited r%d" % branch.revno())
186
# print ("commited r%d" % branch.revno())
193
def _gen_revision_id(branch, when):
194
"""Return new revision-id."""
195
from binascii import hexlify
196
from bzrlib.osutils import rand_bytes, compact_date, user_email
198
s = '%s-%s-' % (user_email(branch), compact_date(when))
199
s += hexlify(rand_bytes(8))
203
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
205
"""Build inventory preparatory to commit.
207
Returns missing_ids, new_inv, any_changes.
209
This adds any changed files into the text store, and sets their
210
test-id, sha and size in the returned inventory appropriately.
213
Modified to hold a list of files that have been deleted from
214
the working directory; these should be removed from the
217
from bzrlib.inventory import Inventory
218
from bzrlib.osutils import isdir, isfile, sha_string, quotefn, \
219
local_time_offset, username, kind_marker, is_inside_any
221
from bzrlib.branch import gen_file_id
222
from bzrlib.errors import BzrError
223
from bzrlib.revision import Revision
224
from bzrlib.trace import mutter, note
227
inv = Inventory(work_inv.root.file_id)
230
for path, entry in work_inv.iter_entries():
231
## TODO: Check that the file kind has not changed from the previous
232
## revision of this file (if any).
234
p = branch.abspath(path)
235
file_id = entry.file_id
236
mutter('commit prep file %s, id %r ' % (p, file_id))
238
if specific_files and not is_inside_any(specific_files, path):
239
mutter(' skipping file excluded from commit')
240
if basis_inv.has_id(file_id):
241
# carry over with previous state
242
inv.add(basis_inv[file_id].copy())
244
# omit this from committed inventory
248
if not work_tree.has_id(file_id):
250
print('deleted %s%s' % (path, kind_marker(entry.kind)))
252
mutter(" file is missing, removing from inventory")
253
missing_ids.append(file_id)
256
# this is present in the new inventory; may be new, modified or
258
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
264
old_kind = old_ie.kind
265
if old_kind != entry.kind:
266
raise BzrError("entry %r changed kind from %r to %r"
267
% (file_id, old_kind, entry.kind))
269
if entry.kind == 'directory':
271
raise BzrError("%s is entered as directory but not a directory"
273
elif entry.kind == 'file':
275
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
277
new_sha1 = work_tree.get_file_sha1(file_id)
280
and old_ie.text_sha1 == new_sha1):
281
## assert content == basis.get_file(file_id).read()
282
entry.text_id = old_ie.text_id
283
entry.text_sha1 = new_sha1
284
entry.text_size = old_ie.text_size
285
mutter(' unchanged from previous text_id {%s}' %
288
content = file(p, 'rb').read()
290
# calculate the sha again, just in case the file contents
291
# changed since we updated the cache
292
entry.text_sha1 = sha_string(content)
293
entry.text_size = len(content)
295
entry.text_id = gen_file_id(entry.name)
296
branch.text_store.add(content, entry.text_id)
297
mutter(' stored with text_id {%s}' % entry.text_id)
300
marked = path + kind_marker(entry.kind)
302
print 'added', marked
304
elif old_ie == entry:
306
elif (old_ie.name == entry.name
307
and old_ie.parent_id == entry.parent_id):
308
print 'modified', marked
311
print 'renamed', marked
313
elif old_ie != entry:
316
return missing_ids, inv, any_changes
added
removed
bzrlib/commit.py
