1
# Copyright (C) 2005-2010 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
18
# The newly committed revision is going to have a shape corresponding
19
# to that of the working tree. Files that are not in the
20
# working tree and that were in the predecessor are reported as
21
# removed --- this can include files that were either removed from the
22
# inventory or deleted in the working tree. If they were only
23
# deleted from disk, they are removed from the working inventory.
25
# We then consider the remaining entries, which will be in the new
26
# version. Directory entries are simply copied across. File entries
27
# must be checked to see if a new version of the file should be
28
# recorded. For each parent revision tree, we check to see what
29
# version of the file was present. If the file was present in at
30
# least one tree, and if it was the same version in all the trees,
31
# then we can just refer to that version. Otherwise, a new version
32
# representing the merger of the file versions must be added.
34
# TODO: Update hashcache before and after - or does the WorkingTree
37
# TODO: Rather than mashing together the ancestry and storing it back,
38
# perhaps the weave should have single method which does it all in one
39
# go, avoiding a lot of redundant work.
41
# TODO: Perhaps give a warning if one of the revisions marked as
42
# merged is already in the ancestry, and then don't record it as a
45
# TODO: If the file is newly merged but unchanged from the version it
46
# merges from, then it should still be reported as newly added
47
# relative to the basis revision.
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
# the rest of the code; add a deprecation of the old name.
59
from bzrlib.branch import Branch
60
from bzrlib.cleanup import OperationWithCleanups
62
from bzrlib.errors import (BzrError, PointlessCommit,
66
from bzrlib.osutils import (get_user_encoding,
68
minimum_path_selection,
71
from bzrlib.trace import mutter, note, is_quiet
72
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
73
from bzrlib import symbol_versioning
74
from bzrlib.urlutils import unescape_for_display
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 ('added', '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('Committing%s', location)
133
def completed(self, revno, rev_id):
134
self._note('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('deleted %s', path)
145
def missing(self, path):
146
self._note('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
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.
170
"""Create a Commit object.
172
:param reporter: the default reporter to use or None to decide later
174
self.reporter = reporter
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):
233
"""Commit working copy as a new revision.
235
:param message: the commit message (it or message_callback is required)
236
:param message_callback: A callback: message = message_callback(cmt_obj)
238
:param timestamp: if not None, seconds-since-epoch for a
239
postdated/predated commit.
241
:param specific_files: If not None, commit only those files. An empty
242
list means 'commit no files'.
244
:param rev_id: If set, use this as the new revision id.
245
Useful for test or import commands that need to tightly
246
control what revisions are assigned. If you duplicate
247
a revision id that exists elsewhere it is your own fault.
248
If null (default), a time/random revision id is generated.
250
:param allow_pointless: If true (default), commit even if nothing
251
has changed and no merges are recorded.
253
:param strict: If true, don't allow a commit if the working tree
254
contains unknown files.
256
:param revprops: Properties for new revision
257
:param local: Perform a local only commit.
258
:param reporter: the reporter to use or None for the default
259
:param verbose: if True and the reporter is not None, report everything
260
:param recursive: If set to 'down', commit in any subtrees that have
261
pending changes of any sort during this commit.
262
:param exclude: None or a list of relative paths to exclude from the
263
commit. Pending changes to excluded files will be ignored by the
266
operation = OperationWithCleanups(self._commit)
267
self.revprops = revprops or {}
268
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
269
self.config = config or self.config
270
return operation.run(
275
specific_files=specific_files,
277
allow_pointless=allow_pointless,
280
working_tree=working_tree,
283
message_callback=message_callback,
286
possible_master_transports=possible_master_transports)
288
def _commit(self, operation, message, timestamp, timezone, committer,
289
specific_files, rev_id, allow_pointless, strict, verbose,
290
working_tree, local, reporter, message_callback, recursive,
291
exclude, possible_master_transports):
292
mutter('preparing to commit')
294
if working_tree is None:
295
raise BzrError("working_tree must be passed into commit().")
297
self.work_tree = working_tree
298
self.branch = self.work_tree.branch
299
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
300
if not self.branch.repository.supports_rich_root():
301
raise errors.RootNotRich()
302
if message_callback is None:
303
if message is not None:
304
if isinstance(message, str):
305
message = message.decode(get_user_encoding())
306
message_callback = lambda x: message
308
raise BzrError("The message or message_callback keyword"
309
" parameter is required for commit().")
311
self.bound_branch = None
312
self.any_entries_deleted = False
313
if exclude is not None:
314
self.exclude = sorted(
315
minimum_path_selection(exclude))
319
self.master_branch = None
320
self.recursive = recursive
322
# self.specific_files is None to indicate no filter, or any iterable to
323
# indicate a filter - [] means no files at all, as per iter_changes.
324
if specific_files is not None:
325
self.specific_files = sorted(
326
minimum_path_selection(specific_files))
328
self.specific_files = None
330
self.allow_pointless = allow_pointless
331
self.message_callback = message_callback
332
self.timestamp = timestamp
333
self.timezone = timezone
334
self.committer = committer
336
self.verbose = verbose
338
self.work_tree.lock_write()
339
operation.add_cleanup(self.work_tree.unlock)
340
self.parents = self.work_tree.get_parent_ids()
341
# We can use record_iter_changes IFF iter_changes is compatible with
342
# the command line parameters, and the repository has fast delta
343
# generation. See bug 347649.
344
self.use_record_iter_changes = (
346
not self.branch.repository._format.supports_tree_reference and
347
(self.branch.repository._format.fast_deltas or
348
len(self.parents) < 2))
349
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
350
operation.add_cleanup(self.pb.finished)
351
self.basis_revid = self.work_tree.last_revision()
352
self.basis_tree = self.work_tree.basis_tree()
353
self.basis_tree.lock_read()
354
operation.add_cleanup(self.basis_tree.unlock)
355
# Cannot commit with conflicts present.
356
if len(self.work_tree.conflicts()) > 0:
357
raise ConflictsInTree
359
# Setup the bound branch variables as needed.
360
self._check_bound_branch(operation, possible_master_transports)
362
# Check that the working tree is up to date
363
old_revno, new_revno = self._check_out_of_date_tree()
365
# Complete configuration setup
366
if reporter is not None:
367
self.reporter = reporter
368
elif self.reporter is None:
369
self.reporter = self._select_reporter()
370
if self.config is None:
371
self.config = self.branch.get_config()
373
self._set_specific_file_ids()
375
# Setup the progress bar. As the number of files that need to be
376
# committed in unknown, progress is reported as stages.
377
# We keep track of entries separately though and include that
378
# information in the progress bar during the relevant stages.
379
self.pb_stage_name = ""
380
self.pb_stage_count = 0
381
self.pb_stage_total = 5
382
if self.bound_branch:
383
self.pb_stage_total += 1
384
self.pb.show_pct = False
385
self.pb.show_spinner = False
386
self.pb.show_eta = False
387
self.pb.show_count = True
388
self.pb.show_bar = True
390
self._gather_parents()
391
# After a merge, a selected file commit is not supported.
392
# See 'bzr help merge' for an explanation as to why.
393
if len(self.parents) > 1 and self.specific_files is not None:
394
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
395
# Excludes are a form of selected file commit.
396
if len(self.parents) > 1 and self.exclude:
397
raise errors.CannotCommitSelectedFileMerge(self.exclude)
399
# Collect the changes
400
self._set_progress_stage("Collecting changes", counter=True)
401
self.builder = self.branch.get_commit_builder(self.parents,
402
self.config, timestamp, timezone, committer, self.revprops, rev_id)
405
self.builder.will_record_deletes()
406
# find the location being committed to
407
if self.bound_branch:
408
master_location = self.master_branch.base
410
master_location = self.branch.base
412
# report the start of the commit
413
self.reporter.started(new_revno, self.rev_id, master_location)
415
self._update_builder_with_changes()
416
self._check_pointless()
418
# TODO: Now the new inventory is known, check for conflicts.
419
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
420
# weave lines, because nothing should be recorded until it is known
421
# that commit will succeed.
422
self._set_progress_stage("Saving data locally")
423
self.builder.finish_inventory()
425
# Prompt the user for a commit message if none provided
426
message = message_callback(self)
427
self.message = message
429
# Add revision data to the local branch
430
self.rev_id = self.builder.commit(self.message)
433
mutter("aborting commit write group because of exception:")
434
trace.log_exception_quietly()
435
note("aborting commit write group: %r" % (e,))
439
self._process_pre_hooks(old_revno, new_revno)
441
# Upload revision data to the master.
442
# this will propagate merged revisions too if needed.
443
if self.bound_branch:
444
self._set_progress_stage("Uploading data to master branch")
445
# 'commit' to the master first so a timeout here causes the
446
# local branch to be out of date
447
self.master_branch.import_last_revision_info(
448
self.branch.repository, new_revno, self.rev_id)
450
# and now do the commit locally.
451
self.branch.set_last_revision_info(new_revno, self.rev_id)
453
# Make the working tree be up to date with the branch. This
454
# includes automatic changes scheduled to be made to the tree, such
455
# as updating its basis and unversioning paths that were missing.
456
self.work_tree.unversion(self.deleted_ids)
457
self._set_progress_stage("Updating the working tree")
458
self.work_tree.update_basis_by_delta(self.rev_id,
459
self.builder.get_basis_delta())
460
self.reporter.completed(new_revno, self.rev_id)
461
self._process_post_hooks(old_revno, new_revno)
464
def _select_reporter(self):
465
"""Select the CommitReporter to use."""
467
return NullCommitReporter()
468
return ReportCommitToLog()
470
def _check_pointless(self):
471
if self.allow_pointless:
473
# A merge with no effect on files
474
if len(self.parents) > 1:
476
# TODO: we could simplify this by using self.builder.basis_delta.
478
# The initial commit adds a root directory, but this in itself is not
479
# a worthwhile commit.
480
if (self.basis_revid == revision.NULL_REVISION and
481
((self.builder.new_inventory is not None and
482
len(self.builder.new_inventory) == 1) or
483
len(self.builder._basis_delta) == 1)):
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):
484
102
raise PointlessCommit()
485
if self.builder.any_changes():
487
raise PointlessCommit()
489
def _check_bound_branch(self, operation, possible_master_transports=None):
490
"""Check to see if the local branch is bound.
492
If it is bound, then most of the commit will actually be
493
done using the remote branch as the target branch.
494
Only at the end will the local branch be updated.
496
if self.local and not self.branch.get_bound_location():
497
raise errors.LocalRequiresBoundBranch()
500
self.master_branch = self.branch.get_master_branch(
501
possible_master_transports)
503
if not self.master_branch:
504
# make this branch the reference branch for out of date checks.
505
self.master_branch = self.branch
508
# If the master branch is bound, we must fail
509
master_bound_location = self.master_branch.get_bound_location()
510
if master_bound_location:
511
raise errors.CommitToDoubleBoundBranch(self.branch,
512
self.master_branch, master_bound_location)
514
# TODO: jam 20051230 We could automatically push local
515
# commits to the remote branch if they would fit.
516
# But for now, just require remote to be identical
519
# Make sure the local branch is identical to the master
520
master_info = self.master_branch.last_revision_info()
521
local_info = self.branch.last_revision_info()
522
if local_info != master_info:
523
raise errors.BoundBranchOutOfDate(self.branch,
526
# Now things are ready to change the master branch
528
self.bound_branch = self.branch
529
self.master_branch.lock_write()
530
operation.add_cleanup(self.master_branch.unlock)
532
def _check_out_of_date_tree(self):
533
"""Check that the working tree is up to date.
535
:return: old_revision_number,new_revision_number tuple
538
first_tree_parent = self.work_tree.get_parent_ids()[0]
540
# if there are no parents, treat our parent as 'None'
541
# this is so that we still consider the master branch
542
# - in a checkout scenario the tree may have no
543
# parents but the branch may do.
544
first_tree_parent = bzrlib.revision.NULL_REVISION
545
old_revno, master_last = self.master_branch.last_revision_info()
546
if master_last != first_tree_parent:
547
if master_last != bzrlib.revision.NULL_REVISION:
548
raise errors.OutOfDateTree(self.work_tree)
549
if self.branch.repository.has_revision(first_tree_parent):
550
new_revno = old_revno + 1
552
# ghost parents never appear in revision history.
554
return old_revno,new_revno
556
def _process_pre_hooks(self, old_revno, new_revno):
557
"""Process any registered pre commit hooks."""
558
self._set_progress_stage("Running pre_commit hooks")
559
self._process_hooks("pre_commit", old_revno, new_revno)
561
def _process_post_hooks(self, old_revno, new_revno):
562
"""Process any registered post commit hooks."""
563
# Process the post commit hooks, if any
564
self._set_progress_stage("Running post_commit hooks")
565
# old style commit hooks - should be deprecated ? (obsoleted in
567
if self.config.post_commit() is not None:
568
hooks = self.config.post_commit().split(' ')
569
# this would be nicer with twisted.python.reflect.namedAny
571
result = eval(hook + '(branch, rev_id)',
572
{'branch':self.branch,
574
'rev_id':self.rev_id})
575
# process new style post commit hooks
576
self._process_hooks("post_commit", old_revno, new_revno)
578
def _process_hooks(self, hook_name, old_revno, new_revno):
579
if not Branch.hooks[hook_name]:
582
# new style commit hooks:
583
if not self.bound_branch:
584
hook_master = self.branch
587
hook_master = self.master_branch
588
hook_local = self.branch
589
# With bound branches, when the master is behind the local branch,
590
# the 'old_revno' and old_revid values here are incorrect.
591
# XXX: FIXME ^. RBC 20060206
593
old_revid = self.parents[0]
595
old_revid = bzrlib.revision.NULL_REVISION
597
if hook_name == "pre_commit":
598
future_tree = self.builder.revision_tree()
599
tree_delta = future_tree.changes_from(self.basis_tree,
602
for hook in Branch.hooks[hook_name]:
603
# show the running hook in the progress bar. As hooks may
604
# end up doing nothing (e.g. because they are not configured by
605
# the user) this is still showing progress, not showing overall
606
# actions - its up to each plugin to show a UI if it want's to
607
# (such as 'Emailing diff to foo@example.com').
608
self.pb_stage_name = "Running %s hooks [%s]" % \
609
(hook_name, Branch.hooks.get_hook_name(hook))
610
self._emit_progress()
611
if 'hooks' in debug.debug_flags:
612
mutter("Invoking commit hook: %r", hook)
613
if hook_name == "post_commit":
614
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
616
elif hook_name == "pre_commit":
617
hook(hook_local, hook_master,
618
old_revno, old_revid, new_revno, self.rev_id,
619
tree_delta, future_tree)
621
def _gather_parents(self):
622
"""Record the parents of a merge for merge detection."""
623
# TODO: Make sure that this list doesn't contain duplicate
624
# entries and the order is preserved when doing this.
625
if self.use_record_iter_changes:
627
self.basis_inv = self.basis_tree.inventory
628
self.parent_invs = [self.basis_inv]
629
for revision in self.parents[1:]:
630
if self.branch.repository.has_revision(revision):
631
mutter('commit parent revision {%s}', revision)
632
inventory = self.branch.repository.get_inventory(revision)
633
self.parent_invs.append(inventory)
635
mutter('commit parent ghost revision {%s}', revision)
637
def _update_builder_with_changes(self):
638
"""Update the commit builder with the data about what has changed.
640
exclude = self.exclude
641
specific_files = self.specific_files
642
mutter("Selecting files for commit with filter %s", specific_files)
645
if self.use_record_iter_changes:
646
iter_changes = self.work_tree.iter_changes(self.basis_tree,
647
specific_files=specific_files)
648
iter_changes = self._filter_iter_changes(iter_changes)
649
for file_id, path, fs_hash in self.builder.record_iter_changes(
650
self.work_tree, self.basis_revid, iter_changes):
651
self.work_tree._observed_sha1(file_id, path, fs_hash)
653
# Build the new inventory
654
self._populate_from_inventory()
655
self._record_unselected()
656
self._report_and_accumulate_deletes()
658
def _filter_iter_changes(self, iter_changes):
659
"""Process iter_changes.
661
This method reports on the changes in iter_changes to the user, and
662
converts 'missing' entries in the iter_changes iterator to 'deleted'
663
entries. 'missing' entries have their
665
:param iter_changes: An iter_changes to process.
666
:return: A generator of changes.
668
reporter = self.reporter
669
report_changes = reporter.is_verbose()
671
for change in iter_changes:
673
old_path = change[1][0]
674
new_path = change[1][1]
675
versioned = change[3][1]
677
versioned = change[3][1]
678
if kind is None and versioned:
681
reporter.missing(new_path)
682
deleted_ids.append(change[0])
683
# Reset the new path (None) and new versioned flag (False)
684
change = (change[0], (change[1][0], None), change[2],
685
(change[3][0], False)) + change[4:]
686
elif kind == 'tree-reference':
687
if self.recursive == 'down':
688
self._commit_nested_tree(change[0], change[1][1])
689
if change[3][0] or change[3][1]:
693
reporter.deleted(old_path)
694
elif old_path is None:
695
reporter.snapshot_change('added', new_path)
696
elif old_path != new_path:
697
reporter.renamed('renamed', old_path, new_path)
700
self.work_tree.branch.repository._format.rich_root_data):
701
# Don't report on changes to '' in non rich root
703
reporter.snapshot_change('modified', new_path)
704
self._next_progress_entry()
705
# Unversion IDs that were found to be deleted
706
self.deleted_ids = deleted_ids
708
def _record_unselected(self):
709
# If specific files are selected, then all un-selected files must be
710
# recorded in their previous state. For more details, see
711
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
712
if self.specific_files or self.exclude:
713
specific_files = self.specific_files or []
714
for path, old_ie in self.basis_inv.iter_entries():
715
if old_ie.file_id in self.builder.new_inventory:
716
# already added - skip.
718
if (is_inside_any(specific_files, path)
719
and not is_inside_any(self.exclude, path)):
720
# was inside the selected path, and not excluded - if not
721
# present it has been deleted so skip.
723
# From here down it was either not selected, or was excluded:
724
# We preserve the entry unaltered.
726
# Note: specific file commits after a merge are currently
727
# prohibited. This test is for sanity/safety in case it's
728
# required after that changes.
729
if len(self.parents) > 1:
731
self.builder.record_entry_contents(ie, self.parent_invs, path,
732
self.basis_tree, None)
734
def _report_and_accumulate_deletes(self):
735
if (isinstance(self.basis_inv, Inventory)
736
and isinstance(self.builder.new_inventory, Inventory)):
737
# the older Inventory classes provide a _byid dict, and building a
738
# set from the keys of this dict is substantially faster than even
739
# getting a set of ids from the inventory
741
# <lifeless> set(dict) is roughly the same speed as
742
# set(iter(dict)) and both are significantly slower than
744
deleted_ids = set(self.basis_inv._byid.keys()) - \
745
set(self.builder.new_inventory._byid.keys())
747
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
749
self.any_entries_deleted = True
750
deleted = [(self.basis_tree.id2path(file_id), file_id)
751
for file_id in deleted_ids]
753
# XXX: this is not quite directory-order sorting
754
for path, file_id in deleted:
755
self.builder.record_delete(path, file_id)
756
self.reporter.deleted(path)
758
def _check_strict(self):
759
# XXX: when we use iter_changes this would likely be faster if
760
# iter_changes would check for us (even in the presence of
763
# raise an exception as soon as we find a single unknown.
764
for unknown in self.work_tree.unknowns():
765
raise StrictCommitFailed()
767
def _populate_from_inventory(self):
768
"""Populate the CommitBuilder by walking the working tree inventory."""
769
# Build the revision inventory.
771
# This starts by creating a new empty inventory. Depending on
772
# which files are selected for commit, and what is present in the
773
# current tree, the new inventory is populated. inventory entries
774
# which are candidates for modification have their revision set to
775
# None; inventory entries that are carried over untouched have their
776
# revision set to their prior value.
778
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
779
# results to create a new inventory at the same time, which results
780
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
783
specific_files = self.specific_files
784
exclude = self.exclude
785
report_changes = self.reporter.is_verbose()
787
# A tree of paths that have been deleted. E.g. if foo/bar has been
788
# deleted, then we have {'foo':{'bar':{}}}
790
# XXX: Note that entries may have the wrong kind because the entry does
791
# not reflect the status on disk.
792
work_inv = self.work_tree.inventory
793
# NB: entries will include entries within the excluded ids/paths
794
# because iter_entries_by_dir has no 'exclude' facility today.
795
entries = work_inv.iter_entries_by_dir(
796
specific_file_ids=self.specific_file_ids, yield_parents=True)
797
for path, existing_ie in entries:
798
file_id = existing_ie.file_id
799
name = existing_ie.name
800
parent_id = existing_ie.parent_id
801
kind = existing_ie.kind
802
# Skip files that have been deleted from the working tree.
803
# The deleted path ids are also recorded so they can be explicitly
806
path_segments = splitpath(path)
807
deleted_dict = deleted_paths
808
for segment in path_segments:
809
deleted_dict = deleted_dict.get(segment, None)
811
# We either took a path not present in the dict
812
# (deleted_dict was None), or we've reached an empty
813
# child dir in the dict, so are now a sub-path.
817
if deleted_dict is not None:
818
# the path has a deleted parent, do not add it.
820
if exclude and is_inside_any(exclude, path):
821
# Skip excluded paths. Excluded paths are processed by
822
# _update_builder_with_changes.
824
content_summary = self.work_tree.path_content_summary(path)
825
kind = content_summary[0]
826
# Note that when a filter of specific files is given, we must only
827
# skip/record deleted files matching that filter.
828
if not specific_files or is_inside_any(specific_files, path):
829
if kind == 'missing':
830
if not deleted_paths:
831
# path won't have been split yet.
832
path_segments = splitpath(path)
833
deleted_dict = deleted_paths
834
for segment in path_segments:
835
deleted_dict = deleted_dict.setdefault(segment, {})
836
self.reporter.missing(path)
837
self._next_progress_entry()
838
deleted_ids.append(file_id)
840
# TODO: have the builder do the nested commit just-in-time IF and
842
if kind == 'tree-reference':
843
# enforce repository nested tree policy.
844
if (not self.work_tree.supports_tree_reference() or
845
# repository does not support it either.
846
not self.branch.repository._format.supports_tree_reference):
848
content_summary = (kind, None, None, None)
849
elif self.recursive == 'down':
850
nested_revision_id = self._commit_nested_tree(
852
content_summary = (kind, None, None, nested_revision_id)
854
nested_revision_id = self.work_tree.get_reference_revision(file_id)
855
content_summary = (kind, None, None, nested_revision_id)
857
# Record an entry for this item
858
# Note: I don't particularly want to have the existing_ie
859
# parameter but the test suite currently (28-Jun-07) breaks
860
# without it thanks to a unicode normalisation issue. :-(
861
definitely_changed = kind != existing_ie.kind
862
self._record_entry(path, file_id, specific_files, kind, name,
863
parent_id, definitely_changed, existing_ie, report_changes,
866
# Unversion IDs that were found to be deleted
867
self.deleted_ids = deleted_ids
869
def _commit_nested_tree(self, file_id, path):
870
"Commit a nested tree."
871
sub_tree = self.work_tree.get_nested_tree(file_id, path)
872
# FIXME: be more comprehensive here:
873
# this works when both trees are in --trees repository,
874
# but when both are bound to a different repository,
875
# it fails; a better way of approaching this is to
876
# finally implement the explicit-caches approach design
877
# a while back - RBC 20070306.
878
if sub_tree.branch.repository.has_same_location(
879
self.work_tree.branch.repository):
880
sub_tree.branch.repository = \
881
self.work_tree.branch.repository
883
return sub_tree.commit(message=None, revprops=self.revprops,
884
recursive=self.recursive,
885
message_callback=self.message_callback,
886
timestamp=self.timestamp, timezone=self.timezone,
887
committer=self.committer,
888
allow_pointless=self.allow_pointless,
889
strict=self.strict, verbose=self.verbose,
890
local=self.local, reporter=self.reporter)
891
except errors.PointlessCommit:
892
return self.work_tree.get_reference_revision(file_id)
894
def _record_entry(self, path, file_id, specific_files, kind, name,
895
parent_id, definitely_changed, existing_ie, report_changes,
897
"Record the new inventory entry for a path if any."
898
# mutter('check %s {%s}', path, file_id)
899
# mutter('%s selected for commit', path)
900
if definitely_changed or existing_ie is None:
901
ie = make_entry(kind, name, parent_id, file_id)
903
ie = existing_ie.copy()
905
# For carried over entries we don't care about the fs hash - the repo
906
# isn't generating a sha, so we're not saving computation time.
907
_, _, fs_hash = self.builder.record_entry_contents(
908
ie, self.parent_invs, path, self.work_tree, content_summary)
910
self._report_change(ie, path)
912
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
915
def _report_change(self, ie, path):
916
"""Report a change to the user.
918
The change that has occurred is described relative to the basis
921
if (self.basis_inv.has_id(ie.file_id)):
922
basis_ie = self.basis_inv[ie.file_id]
925
change = ie.describe_change(basis_ie, ie)
926
if change in (InventoryEntry.RENAMED,
927
InventoryEntry.MODIFIED_AND_RENAMED):
928
old_path = self.basis_inv.id2path(ie.file_id)
929
self.reporter.renamed(change, old_path, path)
930
self._next_progress_entry()
932
if change == 'unchanged':
934
self.reporter.snapshot_change(change, path)
935
self._next_progress_entry()
937
def _set_progress_stage(self, name, counter=False):
938
"""Set the progress stage and emit an update to the progress bar."""
939
self.pb_stage_name = name
940
self.pb_stage_count += 1
942
self.pb_entries_count = 0
944
self.pb_entries_count = None
945
self._emit_progress()
947
def _next_progress_entry(self):
948
"""Emit an update to the progress bar and increment the entry count."""
949
self.pb_entries_count += 1
950
self._emit_progress()
952
def _emit_progress(self):
953
if self.pb_entries_count is not None:
954
text = "%s [%d] - Stage" % (self.pb_stage_name,
955
self.pb_entries_count)
957
text = "%s - Stage" % (self.pb_stage_name, )
958
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
960
def _set_specific_file_ids(self):
961
"""populate self.specific_file_ids if we will use it."""
962
if not self.use_record_iter_changes:
963
# If provided, ensure the specified files are versioned
964
if self.specific_files is not None:
965
# Note: This routine is being called because it raises
966
# PathNotVersionedError as a side effect of finding the IDs. We
967
# later use the ids we found as input to the working tree
968
# inventory iterator, so we only consider those ids rather than
969
# examining the whole tree again.
970
# XXX: Dont we have filter_unversioned to do this more
972
self.specific_file_ids = tree.find_ids_across_trees(
973
self.specific_files, [self.basis_tree, self.work_tree])
975
self.specific_file_ids = None
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
