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
170
"""Create a Commit object.
172
:param reporter: the default reporter to use or None to decide later
174
self.reporter = reporter
184
allow_pointless=True,
192
message_callback=None,
195
possible_master_transports=None):
196
"""Commit working copy as a new revision.
198
:param message: the commit message (it or message_callback is required)
199
:param message_callback: A callback: message = message_callback(cmt_obj)
201
:param timestamp: if not None, seconds-since-epoch for a
202
postdated/predated commit.
204
:param specific_files: If not None, commit only those files. An empty
205
list means 'commit no files'.
207
:param rev_id: If set, use this as the new revision id.
208
Useful for test or import commands that need to tightly
209
control what revisions are assigned. If you duplicate
210
a revision id that exists elsewhere it is your own fault.
211
If null (default), a time/random revision id is generated.
213
:param allow_pointless: If true (default), commit even if nothing
214
has changed and no merges are recorded.
216
:param strict: If true, don't allow a commit if the working tree
217
contains unknown files.
219
:param revprops: Properties for new revision
220
:param local: Perform a local only commit.
221
:param reporter: the reporter to use or None for the default
222
:param verbose: if True and the reporter is not None, report everything
223
:param recursive: If set to 'down', commit in any subtrees that have
224
pending changes of any sort during this commit.
225
:param exclude: None or a list of relative paths to exclude from the
226
commit. Pending changes to excluded files will be ignored by the
229
operation = OperationWithCleanups(self._commit)
230
self.revprops = revprops or {}
231
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
232
self.config = config or self.config
233
return operation.run(
238
specific_files=specific_files,
240
allow_pointless=allow_pointless,
243
working_tree=working_tree,
246
message_callback=message_callback,
249
possible_master_transports=possible_master_transports)
251
def _commit(self, operation, message, timestamp, timezone, committer,
252
specific_files, rev_id, allow_pointless, strict, verbose,
253
working_tree, local, reporter, message_callback, recursive,
254
exclude, possible_master_transports):
255
mutter('preparing to commit')
257
if working_tree is None:
258
raise BzrError("working_tree must be passed into commit().")
260
self.work_tree = working_tree
261
self.branch = self.work_tree.branch
262
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
263
if not self.branch.repository.supports_rich_root():
264
raise errors.RootNotRich()
265
if message_callback is None:
266
if message is not None:
267
if isinstance(message, str):
268
message = message.decode(get_user_encoding())
269
message_callback = lambda x: message
271
raise BzrError("The message or message_callback keyword"
272
" parameter is required for commit().")
274
self.bound_branch = None
275
self.any_entries_deleted = False
276
if exclude is not None:
277
self.exclude = sorted(
278
minimum_path_selection(exclude))
282
self.master_branch = None
283
self.recursive = recursive
285
# self.specific_files is None to indicate no filter, or any iterable to
286
# indicate a filter - [] means no files at all, as per iter_changes.
287
if specific_files is not None:
288
self.specific_files = sorted(
289
minimum_path_selection(specific_files))
291
self.specific_files = None
293
self.allow_pointless = allow_pointless
294
self.message_callback = message_callback
295
self.timestamp = timestamp
296
self.timezone = timezone
297
self.committer = committer
299
self.verbose = verbose
301
self.work_tree.lock_write()
302
operation.add_cleanup(self.work_tree.unlock)
303
self.parents = self.work_tree.get_parent_ids()
304
# We can use record_iter_changes IFF iter_changes is compatible with
305
# the command line parameters, and the repository has fast delta
306
# generation. See bug 347649.
307
self.use_record_iter_changes = (
309
not self.branch.repository._format.supports_tree_reference and
310
(self.branch.repository._format.fast_deltas or
311
len(self.parents) < 2))
312
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
313
operation.add_cleanup(self.pb.finished)
314
self.basis_revid = self.work_tree.last_revision()
315
self.basis_tree = self.work_tree.basis_tree()
316
self.basis_tree.lock_read()
317
operation.add_cleanup(self.basis_tree.unlock)
318
# Cannot commit with conflicts present.
319
if len(self.work_tree.conflicts()) > 0:
320
raise ConflictsInTree
322
# Setup the bound branch variables as needed.
323
self._check_bound_branch(operation, possible_master_transports)
325
# Check that the working tree is up to date
326
old_revno, new_revno = self._check_out_of_date_tree()
328
# Complete configuration setup
329
if reporter is not None:
330
self.reporter = reporter
331
elif self.reporter is None:
332
self.reporter = self._select_reporter()
333
if self.config is None:
334
self.config = self.branch.get_config()
336
self._set_specific_file_ids()
338
# Setup the progress bar. As the number of files that need to be
339
# committed in unknown, progress is reported as stages.
340
# We keep track of entries separately though and include that
341
# information in the progress bar during the relevant stages.
342
self.pb_stage_name = ""
343
self.pb_stage_count = 0
344
self.pb_stage_total = 5
345
if self.bound_branch:
346
self.pb_stage_total += 1
347
self.pb.show_pct = False
348
self.pb.show_spinner = False
349
self.pb.show_eta = False
350
self.pb.show_count = True
351
self.pb.show_bar = True
353
self._gather_parents()
354
# After a merge, a selected file commit is not supported.
355
# See 'bzr help merge' for an explanation as to why.
356
if len(self.parents) > 1 and self.specific_files is not None:
357
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
358
# Excludes are a form of selected file commit.
359
if len(self.parents) > 1 and self.exclude:
360
raise errors.CannotCommitSelectedFileMerge(self.exclude)
362
# Collect the changes
363
self._set_progress_stage("Collecting changes", counter=True)
364
self.builder = self.branch.get_commit_builder(self.parents,
365
self.config, timestamp, timezone, committer, self.revprops, rev_id)
368
self.builder.will_record_deletes()
369
# find the location being committed to
370
if self.bound_branch:
371
master_location = self.master_branch.base
373
master_location = self.branch.base
375
# report the start of the commit
376
self.reporter.started(new_revno, self.rev_id, master_location)
378
self._update_builder_with_changes()
379
self._check_pointless()
381
# TODO: Now the new inventory is known, check for conflicts.
382
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
383
# weave lines, because nothing should be recorded until it is known
384
# that commit will succeed.
385
self._set_progress_stage("Saving data locally")
386
self.builder.finish_inventory()
388
# Prompt the user for a commit message if none provided
389
message = message_callback(self)
390
self.message = message
392
# Add revision data to the local branch
393
self.rev_id = self.builder.commit(self.message)
396
mutter("aborting commit write group because of exception:")
397
trace.log_exception_quietly()
398
note("aborting commit write group: %r" % (e,))
402
self._process_pre_hooks(old_revno, new_revno)
404
# Upload revision data to the master.
405
# this will propagate merged revisions too if needed.
406
if self.bound_branch:
407
self._set_progress_stage("Uploading data to master branch")
408
# 'commit' to the master first so a timeout here causes the
409
# local branch to be out of date
410
self.master_branch.import_last_revision_info(
411
self.branch.repository, new_revno, self.rev_id)
413
# and now do the commit locally.
414
self.branch.set_last_revision_info(new_revno, self.rev_id)
416
# Make the working tree be up to date with the branch. This
417
# includes automatic changes scheduled to be made to the tree, such
418
# as updating its basis and unversioning paths that were missing.
419
self.work_tree.unversion(self.deleted_ids)
420
self._set_progress_stage("Updating the working tree")
421
self.work_tree.update_basis_by_delta(self.rev_id,
422
self.builder.get_basis_delta())
423
self.reporter.completed(new_revno, self.rev_id)
424
self._process_post_hooks(old_revno, new_revno)
427
def _select_reporter(self):
428
"""Select the CommitReporter to use."""
430
return NullCommitReporter()
431
return ReportCommitToLog()
433
def _check_pointless(self):
434
if self.allow_pointless:
436
# A merge with no effect on files
437
if len(self.parents) > 1:
439
# TODO: we could simplify this by using self.builder.basis_delta.
441
# The initial commit adds a root directory, but this in itself is not
442
# a worthwhile commit.
443
if (self.basis_revid == revision.NULL_REVISION and
444
((self.builder.new_inventory is not None and
445
len(self.builder.new_inventory) == 1) or
446
len(self.builder._basis_delta) == 1)):
447
raise PointlessCommit()
448
if self.builder.any_changes():
450
raise PointlessCommit()
452
def _check_bound_branch(self, operation, possible_master_transports=None):
453
"""Check to see if the local branch is bound.
455
If it is bound, then most of the commit will actually be
456
done using the remote branch as the target branch.
457
Only at the end will the local branch be updated.
459
if self.local and not self.branch.get_bound_location():
460
raise errors.LocalRequiresBoundBranch()
463
self.master_branch = self.branch.get_master_branch(
464
possible_master_transports)
466
if not self.master_branch:
467
# make this branch the reference branch for out of date checks.
468
self.master_branch = self.branch
471
# If the master branch is bound, we must fail
472
master_bound_location = self.master_branch.get_bound_location()
473
if master_bound_location:
474
raise errors.CommitToDoubleBoundBranch(self.branch,
475
self.master_branch, master_bound_location)
477
# TODO: jam 20051230 We could automatically push local
478
# commits to the remote branch if they would fit.
479
# But for now, just require remote to be identical
482
# Make sure the local branch is identical to the master
483
master_info = self.master_branch.last_revision_info()
484
local_info = self.branch.last_revision_info()
485
if local_info != master_info:
486
raise errors.BoundBranchOutOfDate(self.branch,
489
# Now things are ready to change the master branch
491
self.bound_branch = self.branch
492
self.master_branch.lock_write()
493
operation.add_cleanup(self.master_branch.unlock)
495
def _check_out_of_date_tree(self):
496
"""Check that the working tree is up to date.
498
:return: old_revision_number,new_revision_number tuple
501
first_tree_parent = self.work_tree.get_parent_ids()[0]
503
# if there are no parents, treat our parent as 'None'
504
# this is so that we still consider the master branch
505
# - in a checkout scenario the tree may have no
506
# parents but the branch may do.
507
first_tree_parent = bzrlib.revision.NULL_REVISION
508
old_revno, master_last = self.master_branch.last_revision_info()
509
if master_last != first_tree_parent:
510
if master_last != bzrlib.revision.NULL_REVISION:
511
raise errors.OutOfDateTree(self.work_tree)
512
if self.branch.repository.has_revision(first_tree_parent):
513
new_revno = old_revno + 1
515
# ghost parents never appear in revision history.
517
return old_revno,new_revno
519
def _process_pre_hooks(self, old_revno, new_revno):
520
"""Process any registered pre commit hooks."""
521
self._set_progress_stage("Running pre_commit hooks")
522
self._process_hooks("pre_commit", old_revno, new_revno)
524
def _process_post_hooks(self, old_revno, new_revno):
525
"""Process any registered post commit hooks."""
526
# Process the post commit hooks, if any
527
self._set_progress_stage("Running post_commit hooks")
528
# old style commit hooks - should be deprecated ? (obsoleted in
530
if self.config.post_commit() is not None:
531
hooks = self.config.post_commit().split(' ')
532
# this would be nicer with twisted.python.reflect.namedAny
534
result = eval(hook + '(branch, rev_id)',
535
{'branch':self.branch,
537
'rev_id':self.rev_id})
538
# process new style post commit hooks
539
self._process_hooks("post_commit", old_revno, new_revno)
541
def _process_hooks(self, hook_name, old_revno, new_revno):
542
if not Branch.hooks[hook_name]:
545
# new style commit hooks:
546
if not self.bound_branch:
547
hook_master = self.branch
550
hook_master = self.master_branch
551
hook_local = self.branch
552
# With bound branches, when the master is behind the local branch,
553
# the 'old_revno' and old_revid values here are incorrect.
554
# XXX: FIXME ^. RBC 20060206
556
old_revid = self.parents[0]
558
old_revid = bzrlib.revision.NULL_REVISION
560
if hook_name == "pre_commit":
561
future_tree = self.builder.revision_tree()
562
tree_delta = future_tree.changes_from(self.basis_tree,
565
for hook in Branch.hooks[hook_name]:
566
# show the running hook in the progress bar. As hooks may
567
# end up doing nothing (e.g. because they are not configured by
568
# the user) this is still showing progress, not showing overall
569
# actions - its up to each plugin to show a UI if it want's to
570
# (such as 'Emailing diff to foo@example.com').
571
self.pb_stage_name = "Running %s hooks [%s]" % \
572
(hook_name, Branch.hooks.get_hook_name(hook))
573
self._emit_progress()
574
if 'hooks' in debug.debug_flags:
575
mutter("Invoking commit hook: %r", hook)
576
if hook_name == "post_commit":
577
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
579
elif hook_name == "pre_commit":
580
hook(hook_local, hook_master,
581
old_revno, old_revid, new_revno, self.rev_id,
582
tree_delta, future_tree)
584
def _gather_parents(self):
585
"""Record the parents of a merge for merge detection."""
586
# TODO: Make sure that this list doesn't contain duplicate
587
# entries and the order is preserved when doing this.
588
if self.use_record_iter_changes:
590
self.basis_inv = self.basis_tree.inventory
591
self.parent_invs = [self.basis_inv]
592
for revision in self.parents[1:]:
593
if self.branch.repository.has_revision(revision):
594
mutter('commit parent revision {%s}', revision)
595
inventory = self.branch.repository.get_inventory(revision)
596
self.parent_invs.append(inventory)
598
mutter('commit parent ghost revision {%s}', revision)
600
def _update_builder_with_changes(self):
601
"""Update the commit builder with the data about what has changed.
603
exclude = self.exclude
604
specific_files = self.specific_files
605
mutter("Selecting files for commit with filter %s", specific_files)
608
if self.use_record_iter_changes:
609
iter_changes = self.work_tree.iter_changes(self.basis_tree,
610
specific_files=specific_files)
611
iter_changes = self._filter_iter_changes(iter_changes)
612
for file_id, path, fs_hash in self.builder.record_iter_changes(
613
self.work_tree, self.basis_revid, iter_changes):
614
self.work_tree._observed_sha1(file_id, path, fs_hash)
616
# Build the new inventory
617
self._populate_from_inventory()
618
self._record_unselected()
619
self._report_and_accumulate_deletes()
621
def _filter_iter_changes(self, iter_changes):
622
"""Process iter_changes.
624
This method reports on the changes in iter_changes to the user, and
625
converts 'missing' entries in the iter_changes iterator to 'deleted'
626
entries. 'missing' entries have their
628
:param iter_changes: An iter_changes to process.
629
:return: A generator of changes.
631
reporter = self.reporter
632
report_changes = reporter.is_verbose()
634
for change in iter_changes:
636
old_path = change[1][0]
637
new_path = change[1][1]
638
versioned = change[3][1]
640
versioned = change[3][1]
641
if kind is None and versioned:
644
reporter.missing(new_path)
645
deleted_ids.append(change[0])
646
# Reset the new path (None) and new versioned flag (False)
647
change = (change[0], (change[1][0], None), change[2],
648
(change[3][0], False)) + change[4:]
649
elif kind == 'tree-reference':
650
if self.recursive == 'down':
651
self._commit_nested_tree(change[0], change[1][1])
652
if change[3][0] or change[3][1]:
656
reporter.deleted(old_path)
657
elif old_path is None:
658
reporter.snapshot_change('added', new_path)
659
elif old_path != new_path:
660
reporter.renamed('renamed', old_path, new_path)
663
self.work_tree.branch.repository._format.rich_root_data):
664
# Don't report on changes to '' in non rich root
666
reporter.snapshot_change('modified', new_path)
667
self._next_progress_entry()
668
# Unversion IDs that were found to be deleted
669
self.deleted_ids = deleted_ids
671
def _record_unselected(self):
672
# If specific files are selected, then all un-selected files must be
673
# recorded in their previous state. For more details, see
674
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
675
if self.specific_files or self.exclude:
676
specific_files = self.specific_files or []
677
for path, old_ie in self.basis_inv.iter_entries():
678
if old_ie.file_id in self.builder.new_inventory:
679
# already added - skip.
681
if (is_inside_any(specific_files, path)
682
and not is_inside_any(self.exclude, path)):
683
# was inside the selected path, and not excluded - if not
684
# present it has been deleted so skip.
686
# From here down it was either not selected, or was excluded:
687
# We preserve the entry unaltered.
689
# Note: specific file commits after a merge are currently
690
# prohibited. This test is for sanity/safety in case it's
691
# required after that changes.
692
if len(self.parents) > 1:
694
self.builder.record_entry_contents(ie, self.parent_invs, path,
695
self.basis_tree, None)
697
def _report_and_accumulate_deletes(self):
698
if (isinstance(self.basis_inv, Inventory)
699
and isinstance(self.builder.new_inventory, Inventory)):
700
# the older Inventory classes provide a _byid dict, and building a
701
# set from the keys of this dict is substantially faster than even
702
# getting a set of ids from the inventory
704
# <lifeless> set(dict) is roughly the same speed as
705
# set(iter(dict)) and both are significantly slower than
707
deleted_ids = set(self.basis_inv._byid.keys()) - \
708
set(self.builder.new_inventory._byid.keys())
710
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
712
self.any_entries_deleted = True
713
deleted = [(self.basis_tree.id2path(file_id), file_id)
714
for file_id in deleted_ids]
716
# XXX: this is not quite directory-order sorting
717
for path, file_id in deleted:
718
self.builder.record_delete(path, file_id)
719
self.reporter.deleted(path)
721
def _check_strict(self):
722
# XXX: when we use iter_changes this would likely be faster if
723
# iter_changes would check for us (even in the presence of
726
# raise an exception as soon as we find a single unknown.
727
for unknown in self.work_tree.unknowns():
728
raise StrictCommitFailed()
730
def _populate_from_inventory(self):
731
"""Populate the CommitBuilder by walking the working tree inventory."""
732
# Build the revision inventory.
734
# This starts by creating a new empty inventory. Depending on
735
# which files are selected for commit, and what is present in the
736
# current tree, the new inventory is populated. inventory entries
737
# which are candidates for modification have their revision set to
738
# None; inventory entries that are carried over untouched have their
739
# revision set to their prior value.
741
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
742
# results to create a new inventory at the same time, which results
743
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
746
specific_files = self.specific_files
747
exclude = self.exclude
748
report_changes = self.reporter.is_verbose()
750
# A tree of paths that have been deleted. E.g. if foo/bar has been
751
# deleted, then we have {'foo':{'bar':{}}}
753
# XXX: Note that entries may have the wrong kind because the entry does
754
# not reflect the status on disk.
755
work_inv = self.work_tree.inventory
756
# NB: entries will include entries within the excluded ids/paths
757
# because iter_entries_by_dir has no 'exclude' facility today.
758
entries = work_inv.iter_entries_by_dir(
759
specific_file_ids=self.specific_file_ids, yield_parents=True)
760
for path, existing_ie in entries:
761
file_id = existing_ie.file_id
762
name = existing_ie.name
763
parent_id = existing_ie.parent_id
764
kind = existing_ie.kind
765
# Skip files that have been deleted from the working tree.
766
# The deleted path ids are also recorded so they can be explicitly
769
path_segments = splitpath(path)
770
deleted_dict = deleted_paths
771
for segment in path_segments:
772
deleted_dict = deleted_dict.get(segment, None)
774
# We either took a path not present in the dict
775
# (deleted_dict was None), or we've reached an empty
776
# child dir in the dict, so are now a sub-path.
780
if deleted_dict is not None:
781
# the path has a deleted parent, do not add it.
783
if exclude and is_inside_any(exclude, path):
784
# Skip excluded paths. Excluded paths are processed by
785
# _update_builder_with_changes.
787
content_summary = self.work_tree.path_content_summary(path)
788
kind = content_summary[0]
789
# Note that when a filter of specific files is given, we must only
790
# skip/record deleted files matching that filter.
791
if not specific_files or is_inside_any(specific_files, path):
792
if kind == 'missing':
793
if not deleted_paths:
794
# path won't have been split yet.
795
path_segments = splitpath(path)
796
deleted_dict = deleted_paths
797
for segment in path_segments:
798
deleted_dict = deleted_dict.setdefault(segment, {})
799
self.reporter.missing(path)
800
self._next_progress_entry()
801
deleted_ids.append(file_id)
803
# TODO: have the builder do the nested commit just-in-time IF and
805
if kind == 'tree-reference':
806
# enforce repository nested tree policy.
807
if (not self.work_tree.supports_tree_reference() or
808
# repository does not support it either.
809
not self.branch.repository._format.supports_tree_reference):
811
content_summary = (kind, None, None, None)
812
elif self.recursive == 'down':
813
nested_revision_id = self._commit_nested_tree(
815
content_summary = (kind, None, None, nested_revision_id)
817
nested_revision_id = self.work_tree.get_reference_revision(file_id)
818
content_summary = (kind, None, None, nested_revision_id)
820
# Record an entry for this item
821
# Note: I don't particularly want to have the existing_ie
822
# parameter but the test suite currently (28-Jun-07) breaks
823
# without it thanks to a unicode normalisation issue. :-(
824
definitely_changed = kind != existing_ie.kind
825
self._record_entry(path, file_id, specific_files, kind, name,
826
parent_id, definitely_changed, existing_ie, report_changes,
829
# Unversion IDs that were found to be deleted
830
self.deleted_ids = deleted_ids
832
def _commit_nested_tree(self, file_id, path):
833
"Commit a nested tree."
834
sub_tree = self.work_tree.get_nested_tree(file_id, path)
835
# FIXME: be more comprehensive here:
836
# this works when both trees are in --trees repository,
837
# but when both are bound to a different repository,
838
# it fails; a better way of approaching this is to
839
# finally implement the explicit-caches approach design
840
# a while back - RBC 20070306.
841
if sub_tree.branch.repository.has_same_location(
842
self.work_tree.branch.repository):
843
sub_tree.branch.repository = \
844
self.work_tree.branch.repository
846
return sub_tree.commit(message=None, revprops=self.revprops,
847
recursive=self.recursive,
848
message_callback=self.message_callback,
849
timestamp=self.timestamp, timezone=self.timezone,
850
committer=self.committer,
851
allow_pointless=self.allow_pointless,
852
strict=self.strict, verbose=self.verbose,
853
local=self.local, reporter=self.reporter)
854
except errors.PointlessCommit:
855
return self.work_tree.get_reference_revision(file_id)
857
def _record_entry(self, path, file_id, specific_files, kind, name,
858
parent_id, definitely_changed, existing_ie, report_changes,
860
"Record the new inventory entry for a path if any."
861
# mutter('check %s {%s}', path, file_id)
862
# mutter('%s selected for commit', path)
863
if definitely_changed or existing_ie is None:
864
ie = make_entry(kind, name, parent_id, file_id)
866
ie = existing_ie.copy()
868
# For carried over entries we don't care about the fs hash - the repo
869
# isn't generating a sha, so we're not saving computation time.
870
_, _, fs_hash = self.builder.record_entry_contents(
871
ie, self.parent_invs, path, self.work_tree, content_summary)
873
self._report_change(ie, path)
875
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
878
def _report_change(self, ie, path):
879
"""Report a change to the user.
881
The change that has occurred is described relative to the basis
884
if (self.basis_inv.has_id(ie.file_id)):
885
basis_ie = self.basis_inv[ie.file_id]
888
change = ie.describe_change(basis_ie, ie)
889
if change in (InventoryEntry.RENAMED,
890
InventoryEntry.MODIFIED_AND_RENAMED):
891
old_path = self.basis_inv.id2path(ie.file_id)
892
self.reporter.renamed(change, old_path, path)
893
self._next_progress_entry()
895
if change == 'unchanged':
897
self.reporter.snapshot_change(change, path)
898
self._next_progress_entry()
900
def _set_progress_stage(self, name, counter=False):
901
"""Set the progress stage and emit an update to the progress bar."""
902
self.pb_stage_name = name
903
self.pb_stage_count += 1
905
self.pb_entries_count = 0
907
self.pb_entries_count = None
908
self._emit_progress()
910
def _next_progress_entry(self):
911
"""Emit an update to the progress bar and increment the entry count."""
912
self.pb_entries_count += 1
913
self._emit_progress()
915
def _emit_progress(self):
916
if self.pb_entries_count is not None:
917
text = "%s [%d] - Stage" % (self.pb_stage_name,
918
self.pb_entries_count)
920
text = "%s - Stage" % (self.pb_stage_name, )
921
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
923
def _set_specific_file_ids(self):
924
"""populate self.specific_file_ids if we will use it."""
925
if not self.use_record_iter_changes:
926
# If provided, ensure the specified files are versioned
927
if self.specific_files is not None:
928
# Note: This routine is being called because it raises
929
# PathNotVersionedError as a side effect of finding the IDs. We
930
# later use the ids we found as input to the working tree
931
# inventory iterator, so we only consider those ids rather than
932
# examining the whole tree again.
933
# XXX: Dont we have filter_unversioned to do this more
935
self.specific_file_ids = tree.find_ids_across_trees(
936
self.specific_files, [self.basis_tree, self.work_tree])
938
self.specific_file_ids = None
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
19
def commit(branch, message,
26
"""Commit working copy as a new revision.
28
The basic approach is to add all the file texts into the
29
store, then the inventory, then make a new revision pointing
30
to that inventory and store that.
32
This is not quite safe if the working copy changes during the
33
commit; for the moment that is simply not allowed. A better
34
approach is to make a temporary copy of the files before
35
computing their hashes, and then add those hashes in turn to
36
the inventory. This should mean at least that there are no
37
broken hash pointers. There is no way we can get a snapshot
38
of the whole directory at an instant. This would also have to
39
be robust against files disappearing, moving, etc. So the
40
whole thing is a bit hard.
42
timestamp -- if not None, seconds-since-epoch for a
43
postdated/predated commit.
46
If true, commit only those files.
49
If set, use this as the new revision id.
50
Useful for test or import commands that need to tightly
51
control what revisions are assigned. If you duplicate
52
a revision id that exists elsewhere it is your own fault.
53
If null (default), a time/random revision id is generated.
58
from bzrlib.osutils import local_time_offset, username
59
from bzrlib.branch import gen_file_id
60
from bzrlib.errors import BzrError
61
from bzrlib.revision import Revision, RevisionReference
62
from bzrlib.trace import mutter, note
67
# First walk over the working inventory; and both update that
68
# and also build a new revision inventory. The revision
69
# inventory needs to hold the text-id, sha1 and size of the
70
# actual file versions committed in the revision. (These are
71
# not present in the working inventory.) We also need to
72
# detect missing/deleted files, and remove them from the
75
work_tree = branch.working_tree()
76
work_inv = work_tree.inventory
77
basis = branch.basis_tree()
78
basis_inv = basis.inventory
81
note('looking for changes...')
83
missing_ids, new_inv = _gather_commit(branch,
90
for file_id in missing_ids:
91
# Any files that have been deleted are now removed from the
92
# working inventory. Files that were not selected for commit
93
# are left as they were in the working inventory and ommitted
94
# from the revision inventory.
96
# have to do this later so we don't mess up the iterator.
97
# since parents may be removed before their children we
100
# FIXME: There's probably a better way to do this; perhaps
101
# the workingtree should know how to filter itbranch.
102
if work_inv.has_id(file_id):
103
del work_inv[file_id]
107
rev_id = _gen_revision_id(time.time())
110
inv_tmp = tempfile.TemporaryFile()
111
new_inv.write_xml(inv_tmp)
113
branch.inventory_store.add(inv_tmp, inv_id)
114
mutter('new inventory_id is {%s}' % inv_id)
116
# We could also just sha hash the inv_tmp file
117
# however, in the case that branch.inventory_store.add()
118
# ever actually does anything special
119
inv_sha1 = branch.get_inventory_sha1(inv_id)
121
branch._write_inventory(work_inv)
123
if timestamp == None:
124
timestamp = time.time()
126
if committer == None:
127
committer = username()
130
timezone = local_time_offset()
132
mutter("building commit log message")
133
rev = Revision(timestamp=timestamp,
138
inventory_sha1=inv_sha1,
141
precursor_id = branch.last_patch()
143
precursor_sha1 = branch.get_revision_sha1(precursor_id)
144
rev.parents = [RevisionReference(precursor_id, precursor_sha1)]
146
rev_tmp = tempfile.TemporaryFile()
147
rev.write_xml(rev_tmp)
149
branch.revision_store.add(rev_tmp, rev_id)
150
mutter("new revision_id is {%s}" % rev_id)
152
## XXX: Everything up to here can simply be orphaned if we abort
153
## the commit; it will leave junk files behind but that doesn't
156
## TODO: Read back the just-generated changeset, and make sure it
157
## applies and recreates the right state.
159
## TODO: Also calculate and store the inventory SHA1
160
mutter("committing patch r%d" % (branch.revno() + 1))
162
branch.append_revision(rev_id)
165
note("commited r%d" % branch.revno())
171
def _gen_revision_id(when):
172
"""Return new revision-id."""
173
from binascii import hexlify
174
from osutils import rand_bytes, compact_date, user_email
176
s = '%s-%s-' % (user_email(), compact_date(when))
177
s += hexlify(rand_bytes(8))
181
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
183
"""Build inventory preparatory to commit.
185
This adds any changed files into the text store, and sets their
186
test-id, sha and size in the returned inventory appropriately.
189
Modified to hold a list of files that have been deleted from
190
the working directory; these should be removed from the
193
from bzrlib.inventory import Inventory
194
from osutils import isdir, isfile, sha_string, quotefn, \
195
local_time_offset, username, kind_marker, is_inside_any
197
from branch import gen_file_id
198
from errors import BzrError
199
from revision import Revision
200
from bzrlib.trace import mutter, note
205
for path, entry in work_inv.iter_entries():
206
## TODO: Check that the file kind has not changed from the previous
207
## revision of this file (if any).
209
p = branch.abspath(path)
210
file_id = entry.file_id
211
mutter('commit prep file %s, id %r ' % (p, file_id))
213
if specific_files and not is_inside_any(specific_files, path):
214
if basis_inv.has_id(file_id):
215
# carry over with previous state
216
inv.add(basis_inv[file_id].copy())
218
# omit this from committed inventory
222
if not work_tree.has_id(file_id):
224
print('deleted %s%s' % (path, kind_marker(entry.kind)))
225
mutter(" file is missing, removing from inventory")
226
missing_ids.append(file_id)
229
# this is present in the new inventory; may be new, modified or
231
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
237
old_kind = old_ie.kind
238
if old_kind != entry.kind:
239
raise BzrError("entry %r changed kind from %r to %r"
240
% (file_id, old_kind, entry.kind))
242
if entry.kind == 'directory':
244
raise BzrError("%s is entered as directory but not a directory"
246
elif entry.kind == 'file':
248
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
250
new_sha1 = work_tree.get_file_sha1(file_id)
253
and old_ie.text_sha1 == new_sha1):
254
## assert content == basis.get_file(file_id).read()
255
entry.text_id = old_ie.text_id
256
entry.text_sha1 = new_sha1
257
entry.text_size = old_ie.text_size
258
mutter(' unchanged from previous text_id {%s}' %
261
content = file(p, 'rb').read()
263
# calculate the sha again, just in case the file contents
264
# changed since we updated the cache
265
entry.text_sha1 = sha_string(content)
266
entry.text_size = len(content)
268
entry.text_id = gen_file_id(entry.name)
269
branch.text_store.add(content, entry.text_id)
270
mutter(' stored with text_id {%s}' % entry.text_id)
273
marked = path + kind_marker(entry.kind)
275
print 'added', marked
276
elif old_ie == entry:
278
elif (old_ie.name == entry.name
279
and old_ie.parent_id == entry.parent_id):
280
print 'modified', marked
282
print 'renamed', marked
284
return missing_ids, inv
added
removed
bzrlib/commit.py
