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.
57
from cStringIO import StringIO
67
from bzrlib.branch import Branch
68
from bzrlib.cleanup import OperationWithCleanups
70
from bzrlib.errors import (BzrError, PointlessCommit,
74
from bzrlib.osutils import (get_user_encoding,
75
kind_marker, isdir,isfile, is_inside_any,
76
is_inside_or_parent_of_any,
77
minimum_path_selection,
78
quotefn, sha_file, split_lines,
81
from bzrlib.testament import Testament
82
from bzrlib.trace import mutter, note, warning, is_quiet
83
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
84
from bzrlib import symbol_versioning
85
from bzrlib.symbol_versioning import (deprecated_passed,
88
from bzrlib.workingtree import WorkingTree
89
from bzrlib.urlutils import unescape_for_display
93
class NullCommitReporter(object):
94
"""I report on progress of a commit."""
96
def started(self, revno, revid, location=None):
98
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
99
"to started.", DeprecationWarning,
103
def snapshot_change(self, change, path):
106
def completed(self, revno, rev_id):
109
def deleted(self, path):
112
def missing(self, path):
115
def renamed(self, change, old_path, new_path):
118
def is_verbose(self):
122
class ReportCommitToLog(NullCommitReporter):
124
def _note(self, format, *args):
127
Subclasses may choose to override this method.
131
def snapshot_change(self, change, path):
132
if path == '' and change in ('added', 'modified'):
134
self._note("%s %s", change, path)
136
def started(self, revno, rev_id, location=None):
137
if location is not None:
138
location = ' to: ' + unescape_for_display(location, 'utf-8')
140
# When started was added, location was only made optional by
141
# accident. Matt Nordhoff 20071129
142
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
143
"to started.", DeprecationWarning,
146
self._note('Committing%s', location)
148
def completed(self, revno, rev_id):
149
self._note('Committed revision %d.', revno)
150
# self._note goes to the console too; so while we want to log the
151
# rev_id, we can't trivially only log it. (See bug 526425). Long
152
# term we should rearrange the reporting structure, but for now
153
# we just mutter seperately. We mutter the revid and revno together
154
# so that concurrent bzr invocations won't lead to confusion.
155
mutter('Committed revid %s as revno %d.', rev_id, revno)
157
def deleted(self, path):
158
self._note('deleted %s', path)
160
def missing(self, path):
161
self._note('missing %s', path)
163
def renamed(self, change, old_path, new_path):
164
self._note('%s %s => %s', change, old_path, new_path)
166
def is_verbose(self):
170
class Commit(object):
171
"""Task of committing a new revision.
173
This is a MethodObject: it accumulates state as the commit is
174
prepared, and then it is discarded. It doesn't represent
175
historical revisions, just the act of recording a new one.
178
Modified to hold a list of files that have been deleted from
179
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.
185
"""Create a Commit object.
187
:param reporter: the default reporter to use or None to decide later
189
self.reporter = reporter
199
allow_pointless=True,
207
message_callback=None,
210
possible_master_transports=None):
211
"""Commit working copy as a new revision.
213
:param message: the commit message (it or message_callback is required)
214
:param message_callback: A callback: message = message_callback(cmt_obj)
216
:param timestamp: if not None, seconds-since-epoch for a
217
postdated/predated commit.
219
:param specific_files: If not None, commit only those files. An empty
220
list means 'commit no files'.
222
:param rev_id: If set, use this as the new revision id.
223
Useful for test or import commands that need to tightly
224
control what revisions are assigned. If you duplicate
225
a revision id that exists elsewhere it is your own fault.
226
If null (default), a time/random revision id is generated.
228
:param allow_pointless: If true (default), commit even if nothing
229
has changed and no merges are recorded.
231
:param strict: If true, don't allow a commit if the working tree
232
contains unknown files.
234
:param revprops: Properties for new revision
235
:param local: Perform a local only commit.
236
:param reporter: the reporter to use or None for the default
237
:param verbose: if True and the reporter is not None, report everything
238
:param recursive: If set to 'down', commit in any subtrees that have
239
pending changes of any sort during this commit.
240
:param exclude: None or a list of relative paths to exclude from the
241
commit. Pending changes to excluded files will be ignored by the
244
operation = OperationWithCleanups(self._commit)
245
self.revprops = revprops or {}
246
# XXX: Can be set on __init__ or passed in - this is a bit ugly.
247
self.config = config or self.config
248
return operation.run(
253
specific_files=specific_files,
255
allow_pointless=allow_pointless,
258
working_tree=working_tree,
261
message_callback=message_callback,
264
possible_master_transports=possible_master_transports)
266
def _commit(self, operation, message, timestamp, timezone, committer,
267
specific_files, rev_id, allow_pointless, strict, verbose,
268
working_tree, local, reporter, message_callback, recursive,
269
exclude, possible_master_transports):
270
mutter('preparing to commit')
272
if working_tree is None:
273
raise BzrError("working_tree must be passed into commit().")
275
self.work_tree = working_tree
276
self.branch = self.work_tree.branch
277
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
278
if not self.branch.repository.supports_rich_root():
279
raise errors.RootNotRich()
280
if message_callback is None:
281
if message is not None:
282
if isinstance(message, str):
283
message = message.decode(get_user_encoding())
284
message_callback = lambda x: message
286
raise BzrError("The message or message_callback keyword"
287
" parameter is required for commit().")
289
self.bound_branch = None
290
self.any_entries_deleted = False
291
if exclude is not None:
292
self.exclude = sorted(
293
minimum_path_selection(exclude))
297
self.master_branch = None
298
self.recursive = recursive
300
# self.specific_files is None to indicate no filter, or any iterable to
301
# indicate a filter - [] means no files at all, as per iter_changes.
302
if specific_files is not None:
303
self.specific_files = sorted(
304
minimum_path_selection(specific_files))
306
self.specific_files = None
308
self.allow_pointless = allow_pointless
309
self.message_callback = message_callback
310
self.timestamp = timestamp
311
self.timezone = timezone
312
self.committer = committer
314
self.verbose = verbose
316
self.work_tree.lock_write()
317
operation.add_cleanup(self.work_tree.unlock)
318
self.parents = self.work_tree.get_parent_ids()
319
# We can use record_iter_changes IFF iter_changes is compatible with
320
# the command line parameters, and the repository has fast delta
321
# generation. See bug 347649.
322
self.use_record_iter_changes = (
324
not self.branch.repository._format.supports_tree_reference and
325
(self.branch.repository._format.fast_deltas or
326
len(self.parents) < 2))
327
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
328
operation.add_cleanup(self.pb.finished)
329
self.basis_revid = self.work_tree.last_revision()
330
self.basis_tree = self.work_tree.basis_tree()
331
self.basis_tree.lock_read()
332
operation.add_cleanup(self.basis_tree.unlock)
333
# Cannot commit with conflicts present.
334
if len(self.work_tree.conflicts()) > 0:
335
raise ConflictsInTree
337
# Setup the bound branch variables as needed.
338
self._check_bound_branch(operation, possible_master_transports)
340
# Check that the working tree is up to date
341
old_revno, new_revno = self._check_out_of_date_tree()
343
# Complete configuration setup
344
if reporter is not None:
345
self.reporter = reporter
346
elif self.reporter is None:
347
self.reporter = self._select_reporter()
348
if self.config is None:
349
self.config = self.branch.get_config()
351
self._set_specific_file_ids()
353
# Setup the progress bar. As the number of files that need to be
354
# committed in unknown, progress is reported as stages.
355
# We keep track of entries separately though and include that
356
# information in the progress bar during the relevant stages.
357
self.pb_stage_name = ""
358
self.pb_stage_count = 0
359
self.pb_stage_total = 5
360
if self.bound_branch:
361
self.pb_stage_total += 1
362
self.pb.show_pct = False
363
self.pb.show_spinner = False
364
self.pb.show_eta = False
365
self.pb.show_count = True
366
self.pb.show_bar = True
368
self._gather_parents()
369
# After a merge, a selected file commit is not supported.
370
# See 'bzr help merge' for an explanation as to why.
371
if len(self.parents) > 1 and self.specific_files is not None:
372
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
373
# Excludes are a form of selected file commit.
374
if len(self.parents) > 1 and self.exclude:
375
raise errors.CannotCommitSelectedFileMerge(self.exclude)
377
# Collect the changes
378
self._set_progress_stage("Collecting changes", counter=True)
379
self.builder = self.branch.get_commit_builder(self.parents,
380
self.config, timestamp, timezone, committer, self.revprops, rev_id)
383
self.builder.will_record_deletes()
384
# find the location being committed to
385
if self.bound_branch:
386
master_location = self.master_branch.base
388
master_location = self.branch.base
390
# report the start of the commit
391
self.reporter.started(new_revno, self.rev_id, master_location)
393
self._update_builder_with_changes()
394
self._check_pointless()
396
# TODO: Now the new inventory is known, check for conflicts.
397
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
398
# weave lines, because nothing should be recorded until it is known
399
# that commit will succeed.
400
self._set_progress_stage("Saving data locally")
401
self.builder.finish_inventory()
403
# Prompt the user for a commit message if none provided
404
message = message_callback(self)
405
self.message = message
407
# Add revision data to the local branch
408
self.rev_id = self.builder.commit(self.message)
411
mutter("aborting commit write group because of exception:")
412
trace.log_exception_quietly()
413
note("aborting commit write group: %r" % (e,))
417
self._process_pre_hooks(old_revno, new_revno)
419
# Upload revision data to the master.
420
# this will propagate merged revisions too if needed.
421
if self.bound_branch:
422
self._set_progress_stage("Uploading data to master branch")
423
# 'commit' to the master first so a timeout here causes the
424
# local branch to be out of date
425
self.master_branch.import_last_revision_info(
426
self.branch.repository, new_revno, self.rev_id)
428
# and now do the commit locally.
429
self.branch.set_last_revision_info(new_revno, self.rev_id)
431
# Make the working tree be up to date with the branch. This
432
# includes automatic changes scheduled to be made to the tree, such
433
# as updating its basis and unversioning paths that were missing.
434
self.work_tree.unversion(self.deleted_ids)
435
self._set_progress_stage("Updating the working tree")
436
self.work_tree.update_basis_by_delta(self.rev_id,
437
self.builder.get_basis_delta())
438
self.reporter.completed(new_revno, self.rev_id)
439
self._process_post_hooks(old_revno, new_revno)
442
def _select_reporter(self):
443
"""Select the CommitReporter to use."""
445
return NullCommitReporter()
446
return ReportCommitToLog()
448
def _check_pointless(self):
449
if self.allow_pointless:
451
# A merge with no effect on files
452
if len(self.parents) > 1:
454
# TODO: we could simplify this by using self.builder.basis_delta.
456
# The initial commit adds a root directory, but this in itself is not
457
# a worthwhile commit.
458
if (self.basis_revid == revision.NULL_REVISION and
459
((self.builder.new_inventory is not None and
460
len(self.builder.new_inventory) == 1) or
461
len(self.builder._basis_delta) == 1)):
60
import time, tempfile, re
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):
462
102
raise PointlessCommit()
463
if self.builder.any_changes():
465
raise PointlessCommit()
467
def _check_bound_branch(self, operation, possible_master_transports=None):
468
"""Check to see if the local branch is bound.
470
If it is bound, then most of the commit will actually be
471
done using the remote branch as the target branch.
472
Only at the end will the local branch be updated.
474
if self.local and not self.branch.get_bound_location():
475
raise errors.LocalRequiresBoundBranch()
478
self.master_branch = self.branch.get_master_branch(
479
possible_master_transports)
481
if not self.master_branch:
482
# make this branch the reference branch for out of date checks.
483
self.master_branch = self.branch
486
# If the master branch is bound, we must fail
487
master_bound_location = self.master_branch.get_bound_location()
488
if master_bound_location:
489
raise errors.CommitToDoubleBoundBranch(self.branch,
490
self.master_branch, master_bound_location)
492
# TODO: jam 20051230 We could automatically push local
493
# commits to the remote branch if they would fit.
494
# But for now, just require remote to be identical
497
# Make sure the local branch is identical to the master
498
master_info = self.master_branch.last_revision_info()
499
local_info = self.branch.last_revision_info()
500
if local_info != master_info:
501
raise errors.BoundBranchOutOfDate(self.branch,
504
# Now things are ready to change the master branch
506
self.bound_branch = self.branch
507
self.master_branch.lock_write()
508
operation.add_cleanup(self.master_branch.unlock)
510
def _check_out_of_date_tree(self):
511
"""Check that the working tree is up to date.
513
:return: old_revision_number,new_revision_number tuple
516
first_tree_parent = self.work_tree.get_parent_ids()[0]
518
# if there are no parents, treat our parent as 'None'
519
# this is so that we still consider the master branch
520
# - in a checkout scenario the tree may have no
521
# parents but the branch may do.
522
first_tree_parent = bzrlib.revision.NULL_REVISION
523
old_revno, master_last = self.master_branch.last_revision_info()
524
if master_last != first_tree_parent:
525
if master_last != bzrlib.revision.NULL_REVISION:
526
raise errors.OutOfDateTree(self.work_tree)
527
if self.branch.repository.has_revision(first_tree_parent):
528
new_revno = old_revno + 1
530
# ghost parents never appear in revision history.
532
return old_revno,new_revno
534
def _process_pre_hooks(self, old_revno, new_revno):
535
"""Process any registered pre commit hooks."""
536
self._set_progress_stage("Running pre_commit hooks")
537
self._process_hooks("pre_commit", old_revno, new_revno)
539
def _process_post_hooks(self, old_revno, new_revno):
540
"""Process any registered post commit hooks."""
541
# Process the post commit hooks, if any
542
self._set_progress_stage("Running post_commit hooks")
543
# old style commit hooks - should be deprecated ? (obsoleted in
545
if self.config.post_commit() is not None:
546
hooks = self.config.post_commit().split(' ')
547
# this would be nicer with twisted.python.reflect.namedAny
549
result = eval(hook + '(branch, rev_id)',
550
{'branch':self.branch,
552
'rev_id':self.rev_id})
553
# process new style post commit hooks
554
self._process_hooks("post_commit", old_revno, new_revno)
556
def _process_hooks(self, hook_name, old_revno, new_revno):
557
if not Branch.hooks[hook_name]:
560
# new style commit hooks:
561
if not self.bound_branch:
562
hook_master = self.branch
565
hook_master = self.master_branch
566
hook_local = self.branch
567
# With bound branches, when the master is behind the local branch,
568
# the 'old_revno' and old_revid values here are incorrect.
569
# XXX: FIXME ^. RBC 20060206
571
old_revid = self.parents[0]
573
old_revid = bzrlib.revision.NULL_REVISION
575
if hook_name == "pre_commit":
576
future_tree = self.builder.revision_tree()
577
tree_delta = future_tree.changes_from(self.basis_tree,
580
for hook in Branch.hooks[hook_name]:
581
# show the running hook in the progress bar. As hooks may
582
# end up doing nothing (e.g. because they are not configured by
583
# the user) this is still showing progress, not showing overall
584
# actions - its up to each plugin to show a UI if it want's to
585
# (such as 'Emailing diff to foo@example.com').
586
self.pb_stage_name = "Running %s hooks [%s]" % \
587
(hook_name, Branch.hooks.get_hook_name(hook))
588
self._emit_progress()
589
if 'hooks' in debug.debug_flags:
590
mutter("Invoking commit hook: %r", hook)
591
if hook_name == "post_commit":
592
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
594
elif hook_name == "pre_commit":
595
hook(hook_local, hook_master,
596
old_revno, old_revid, new_revno, self.rev_id,
597
tree_delta, future_tree)
599
def _gather_parents(self):
600
"""Record the parents of a merge for merge detection."""
601
# TODO: Make sure that this list doesn't contain duplicate
602
# entries and the order is preserved when doing this.
603
if self.use_record_iter_changes:
605
self.basis_inv = self.basis_tree.inventory
606
self.parent_invs = [self.basis_inv]
607
for revision in self.parents[1:]:
608
if self.branch.repository.has_revision(revision):
609
mutter('commit parent revision {%s}', revision)
610
inventory = self.branch.repository.get_inventory(revision)
611
self.parent_invs.append(inventory)
613
mutter('commit parent ghost revision {%s}', revision)
615
def _update_builder_with_changes(self):
616
"""Update the commit builder with the data about what has changed.
618
exclude = self.exclude
619
specific_files = self.specific_files
620
mutter("Selecting files for commit with filter %s", specific_files)
623
if self.use_record_iter_changes:
624
iter_changes = self.work_tree.iter_changes(self.basis_tree,
625
specific_files=specific_files)
626
iter_changes = self._filter_iter_changes(iter_changes)
627
for file_id, path, fs_hash in self.builder.record_iter_changes(
628
self.work_tree, self.basis_revid, iter_changes):
629
self.work_tree._observed_sha1(file_id, path, fs_hash)
631
# Build the new inventory
632
self._populate_from_inventory()
633
self._record_unselected()
634
self._report_and_accumulate_deletes()
636
def _filter_iter_changes(self, iter_changes):
637
"""Process iter_changes.
639
This method reports on the changes in iter_changes to the user, and
640
converts 'missing' entries in the iter_changes iterator to 'deleted'
641
entries. 'missing' entries have their
643
:param iter_changes: An iter_changes to process.
644
:return: A generator of changes.
646
reporter = self.reporter
647
report_changes = reporter.is_verbose()
649
for change in iter_changes:
651
old_path = change[1][0]
652
new_path = change[1][1]
653
versioned = change[3][1]
655
versioned = change[3][1]
656
if kind is None and versioned:
659
reporter.missing(new_path)
660
deleted_ids.append(change[0])
661
# Reset the new path (None) and new versioned flag (False)
662
change = (change[0], (change[1][0], None), change[2],
663
(change[3][0], False)) + change[4:]
664
elif kind == 'tree-reference':
665
if self.recursive == 'down':
666
self._commit_nested_tree(change[0], change[1][1])
667
if change[3][0] or change[3][1]:
671
reporter.deleted(old_path)
672
elif old_path is None:
673
reporter.snapshot_change('added', new_path)
674
elif old_path != new_path:
675
reporter.renamed('renamed', old_path, new_path)
678
self.work_tree.branch.repository._format.rich_root_data):
679
# Don't report on changes to '' in non rich root
681
reporter.snapshot_change('modified', new_path)
682
self._next_progress_entry()
683
# Unversion IDs that were found to be deleted
684
self.deleted_ids = deleted_ids
686
def _record_unselected(self):
687
# If specific files are selected, then all un-selected files must be
688
# recorded in their previous state. For more details, see
689
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
690
if self.specific_files or self.exclude:
691
specific_files = self.specific_files or []
692
for path, old_ie in self.basis_inv.iter_entries():
693
if old_ie.file_id in self.builder.new_inventory:
694
# already added - skip.
696
if (is_inside_any(specific_files, path)
697
and not is_inside_any(self.exclude, path)):
698
# was inside the selected path, and not excluded - if not
699
# present it has been deleted so skip.
701
# From here down it was either not selected, or was excluded:
702
# We preserve the entry unaltered.
704
# Note: specific file commits after a merge are currently
705
# prohibited. This test is for sanity/safety in case it's
706
# required after that changes.
707
if len(self.parents) > 1:
709
self.builder.record_entry_contents(ie, self.parent_invs, path,
710
self.basis_tree, None)
712
def _report_and_accumulate_deletes(self):
713
if (isinstance(self.basis_inv, Inventory)
714
and isinstance(self.builder.new_inventory, Inventory)):
715
# the older Inventory classes provide a _byid dict, and building a
716
# set from the keys of this dict is substantially faster than even
717
# getting a set of ids from the inventory
719
# <lifeless> set(dict) is roughly the same speed as
720
# set(iter(dict)) and both are significantly slower than
722
deleted_ids = set(self.basis_inv._byid.keys()) - \
723
set(self.builder.new_inventory._byid.keys())
725
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
727
self.any_entries_deleted = True
728
deleted = [(self.basis_tree.id2path(file_id), file_id)
729
for file_id in deleted_ids]
731
# XXX: this is not quite directory-order sorting
732
for path, file_id in deleted:
733
self.builder.record_delete(path, file_id)
734
self.reporter.deleted(path)
736
def _check_strict(self):
737
# XXX: when we use iter_changes this would likely be faster if
738
# iter_changes would check for us (even in the presence of
741
# raise an exception as soon as we find a single unknown.
742
for unknown in self.work_tree.unknowns():
743
raise StrictCommitFailed()
745
def _populate_from_inventory(self):
746
"""Populate the CommitBuilder by walking the working tree inventory."""
747
# Build the revision inventory.
749
# This starts by creating a new empty inventory. Depending on
750
# which files are selected for commit, and what is present in the
751
# current tree, the new inventory is populated. inventory entries
752
# which are candidates for modification have their revision set to
753
# None; inventory entries that are carried over untouched have their
754
# revision set to their prior value.
756
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
757
# results to create a new inventory at the same time, which results
758
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
761
specific_files = self.specific_files
762
exclude = self.exclude
763
report_changes = self.reporter.is_verbose()
765
# A tree of paths that have been deleted. E.g. if foo/bar has been
766
# deleted, then we have {'foo':{'bar':{}}}
768
# XXX: Note that entries may have the wrong kind because the entry does
769
# not reflect the status on disk.
770
work_inv = self.work_tree.inventory
771
# NB: entries will include entries within the excluded ids/paths
772
# because iter_entries_by_dir has no 'exclude' facility today.
773
entries = work_inv.iter_entries_by_dir(
774
specific_file_ids=self.specific_file_ids, yield_parents=True)
775
for path, existing_ie in entries:
776
file_id = existing_ie.file_id
777
name = existing_ie.name
778
parent_id = existing_ie.parent_id
779
kind = existing_ie.kind
780
# Skip files that have been deleted from the working tree.
781
# The deleted path ids are also recorded so they can be explicitly
784
path_segments = splitpath(path)
785
deleted_dict = deleted_paths
786
for segment in path_segments:
787
deleted_dict = deleted_dict.get(segment, None)
789
# We either took a path not present in the dict
790
# (deleted_dict was None), or we've reached an empty
791
# child dir in the dict, so are now a sub-path.
795
if deleted_dict is not None:
796
# the path has a deleted parent, do not add it.
798
if exclude and is_inside_any(exclude, path):
799
# Skip excluded paths. Excluded paths are processed by
800
# _update_builder_with_changes.
802
content_summary = self.work_tree.path_content_summary(path)
803
kind = content_summary[0]
804
# Note that when a filter of specific files is given, we must only
805
# skip/record deleted files matching that filter.
806
if not specific_files or is_inside_any(specific_files, path):
807
if kind == 'missing':
808
if not deleted_paths:
809
# path won't have been split yet.
810
path_segments = splitpath(path)
811
deleted_dict = deleted_paths
812
for segment in path_segments:
813
deleted_dict = deleted_dict.setdefault(segment, {})
814
self.reporter.missing(path)
815
self._next_progress_entry()
816
deleted_ids.append(file_id)
818
# TODO: have the builder do the nested commit just-in-time IF and
820
if kind == 'tree-reference':
821
# enforce repository nested tree policy.
822
if (not self.work_tree.supports_tree_reference() or
823
# repository does not support it either.
824
not self.branch.repository._format.supports_tree_reference):
826
content_summary = (kind, None, None, None)
827
elif self.recursive == 'down':
828
nested_revision_id = self._commit_nested_tree(
830
content_summary = (kind, None, None, nested_revision_id)
832
nested_revision_id = self.work_tree.get_reference_revision(file_id)
833
content_summary = (kind, None, None, nested_revision_id)
835
# Record an entry for this item
836
# Note: I don't particularly want to have the existing_ie
837
# parameter but the test suite currently (28-Jun-07) breaks
838
# without it thanks to a unicode normalisation issue. :-(
839
definitely_changed = kind != existing_ie.kind
840
self._record_entry(path, file_id, specific_files, kind, name,
841
parent_id, definitely_changed, existing_ie, report_changes,
844
# Unversion IDs that were found to be deleted
845
self.deleted_ids = deleted_ids
847
def _commit_nested_tree(self, file_id, path):
848
"Commit a nested tree."
849
sub_tree = self.work_tree.get_nested_tree(file_id, path)
850
# FIXME: be more comprehensive here:
851
# this works when both trees are in --trees repository,
852
# but when both are bound to a different repository,
853
# it fails; a better way of approaching this is to
854
# finally implement the explicit-caches approach design
855
# a while back - RBC 20070306.
856
if sub_tree.branch.repository.has_same_location(
857
self.work_tree.branch.repository):
858
sub_tree.branch.repository = \
859
self.work_tree.branch.repository
861
return sub_tree.commit(message=None, revprops=self.revprops,
862
recursive=self.recursive,
863
message_callback=self.message_callback,
864
timestamp=self.timestamp, timezone=self.timezone,
865
committer=self.committer,
866
allow_pointless=self.allow_pointless,
867
strict=self.strict, verbose=self.verbose,
868
local=self.local, reporter=self.reporter)
869
except errors.PointlessCommit:
870
return self.work_tree.get_reference_revision(file_id)
872
def _record_entry(self, path, file_id, specific_files, kind, name,
873
parent_id, definitely_changed, existing_ie, report_changes,
875
"Record the new inventory entry for a path if any."
876
# mutter('check %s {%s}', path, file_id)
877
# mutter('%s selected for commit', path)
878
if definitely_changed or existing_ie is None:
879
ie = make_entry(kind, name, parent_id, file_id)
881
ie = existing_ie.copy()
883
# For carried over entries we don't care about the fs hash - the repo
884
# isn't generating a sha, so we're not saving computation time.
885
_, _, fs_hash = self.builder.record_entry_contents(
886
ie, self.parent_invs, path, self.work_tree, content_summary)
888
self._report_change(ie, path)
890
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
893
def _report_change(self, ie, path):
894
"""Report a change to the user.
896
The change that has occurred is described relative to the basis
899
if (self.basis_inv.has_id(ie.file_id)):
900
basis_ie = self.basis_inv[ie.file_id]
903
change = ie.describe_change(basis_ie, ie)
904
if change in (InventoryEntry.RENAMED,
905
InventoryEntry.MODIFIED_AND_RENAMED):
906
old_path = self.basis_inv.id2path(ie.file_id)
907
self.reporter.renamed(change, old_path, path)
908
self._next_progress_entry()
910
if change == 'unchanged':
912
self.reporter.snapshot_change(change, path)
913
self._next_progress_entry()
915
def _set_progress_stage(self, name, counter=False):
916
"""Set the progress stage and emit an update to the progress bar."""
917
self.pb_stage_name = name
918
self.pb_stage_count += 1
920
self.pb_entries_count = 0
922
self.pb_entries_count = None
923
self._emit_progress()
925
def _next_progress_entry(self):
926
"""Emit an update to the progress bar and increment the entry count."""
927
self.pb_entries_count += 1
928
self._emit_progress()
930
def _emit_progress(self):
931
if self.pb_entries_count is not None:
932
text = "%s [%d] - Stage" % (self.pb_stage_name,
933
self.pb_entries_count)
935
text = "%s - Stage" % (self.pb_stage_name, )
936
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
938
def _set_specific_file_ids(self):
939
"""populate self.specific_file_ids if we will use it."""
940
if not self.use_record_iter_changes:
941
# If provided, ensure the specified files are versioned
942
if self.specific_files is not None:
943
# Note: This routine is being called because it raises
944
# PathNotVersionedError as a side effect of finding the IDs. We
945
# later use the ids we found as input to the working tree
946
# inventory iterator, so we only consider those ids rather than
947
# examining the whole tree again.
948
# XXX: Dont we have filter_unversioned to do this more
950
self.specific_file_ids = tree.find_ids_across_trees(
951
self.specific_files, [self.basis_tree, self.work_tree])
953
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
# Python strings can include characters that can't be
148
# represented in well-formed XML; escape characters that
149
# aren't listed in the XML specification
150
# (http://www.w3.org/TR/REC-xml/#NT-Char).
151
if isinstance(message, unicode):
152
char_pattern = u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]'
154
# Use a regular 'str' as pattern to avoid having re.subn
155
# return 'unicode' results.
156
char_pattern = '[^x09\x0A\x0D\x20-\xFF]'
157
message, escape_count = re.subn(
159
lambda match: match.group(0).encode('unicode_escape'),
162
note("replaced %d control characters in message", escape_count)
163
rev = Revision(timestamp=timestamp,
168
inventory_sha1=inv_sha1,
172
precursor_id = branch.last_patch()
174
precursor_sha1 = branch.get_revision_sha1(precursor_id)
175
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
176
for merge_rev in pending_merges:
177
rev.parents.append(RevisionReference(merge_rev))
179
rev_tmp = tempfile.TemporaryFile()
180
serializer_v4.write_revision(rev, rev_tmp)
182
branch.revision_store.add(rev_tmp, rev_id)
183
mutter("new revision_id is {%s}" % rev_id)
185
## XXX: Everything up to here can simply be orphaned if we abort
186
## the commit; it will leave junk files behind but that doesn't
189
## TODO: Read back the just-generated changeset, and make sure it
190
## applies and recreates the right state.
192
## TODO: Also calculate and store the inventory SHA1
193
mutter("committing patch r%d" % (branch.revno() + 1))
195
branch.append_revision(rev_id)
197
branch.set_pending_merges([])
200
# disabled; should go through logging
201
# note("commited r%d" % branch.revno())
202
# print ("commited r%d" % branch.revno())
209
def _gen_revision_id(branch, when):
210
"""Return new revision-id."""
211
from binascii import hexlify
212
from bzrlib.osutils import rand_bytes, compact_date, user_email
214
s = '%s-%s-' % (user_email(branch), compact_date(when))
215
s += hexlify(rand_bytes(8))
219
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
221
"""Build inventory preparatory to commit.
223
Returns missing_ids, new_inv, any_changes.
225
This adds any changed files into the text store, and sets their
226
test-id, sha and size in the returned inventory appropriately.
229
Modified to hold a list of files that have been deleted from
230
the working directory; these should be removed from the
233
from bzrlib.inventory import Inventory
234
from bzrlib.osutils import isdir, isfile, sha_string, quotefn, \
235
local_time_offset, username, kind_marker, is_inside_any
237
from bzrlib.branch import gen_file_id
238
from bzrlib.errors import BzrError
239
from bzrlib.revision import Revision
240
from bzrlib.trace import mutter, note
243
inv = Inventory(work_inv.root.file_id)
246
for path, entry in work_inv.iter_entries():
247
## TODO: Check that the file kind has not changed from the previous
248
## revision of this file (if any).
250
p = branch.abspath(path)
251
file_id = entry.file_id
252
mutter('commit prep file %s, id %r ' % (p, file_id))
254
if specific_files and not is_inside_any(specific_files, path):
255
mutter(' skipping file excluded from commit')
256
if basis_inv.has_id(file_id):
257
# carry over with previous state
258
inv.add(basis_inv[file_id].copy())
260
# omit this from committed inventory
264
if not work_tree.has_id(file_id):
266
print('deleted %s%s' % (path, kind_marker(entry.kind)))
268
mutter(" file is missing, removing from inventory")
269
missing_ids.append(file_id)
272
# this is present in the new inventory; may be new, modified or
274
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
280
old_kind = old_ie.kind
281
if old_kind != entry.kind:
282
raise BzrError("entry %r changed kind from %r to %r"
283
% (file_id, old_kind, entry.kind))
285
if entry.kind == 'directory':
287
raise BzrError("%s is entered as directory but not a directory"
289
elif entry.kind == 'file':
291
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
293
new_sha1 = work_tree.get_file_sha1(file_id)
296
and old_ie.text_sha1 == new_sha1):
297
## assert content == basis.get_file(file_id).read()
298
entry.text_id = old_ie.text_id
299
entry.text_sha1 = new_sha1
300
entry.text_size = old_ie.text_size
301
mutter(' unchanged from previous text_id {%s}' %
304
content = file(p, 'rb').read()
306
# calculate the sha again, just in case the file contents
307
# changed since we updated the cache
308
entry.text_sha1 = sha_string(content)
309
entry.text_size = len(content)
311
entry.text_id = gen_file_id(entry.name)
312
branch.text_store.add(content, entry.text_id)
313
mutter(' stored with text_id {%s}' % entry.text_id)
316
marked = path + kind_marker(entry.kind)
318
print 'added', marked
320
elif old_ie == entry:
322
elif (old_ie.name == entry.name
323
and old_ie.parent_id == entry.parent_id):
324
print 'modified', marked
327
print 'renamed', marked
329
elif old_ie != entry:
332
return missing_ids, inv, any_changes
added
removed
bzrlib/commit.py
