1
# Copyright (C) 2005, 2006, 2007 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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 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
65
from bzrlib.branch import Branch
67
from bzrlib.errors import (BzrError, PointlessCommit,
71
from bzrlib.osutils import (kind_marker, isdir,isfile, is_inside_any,
72
is_inside_or_parent_of_any,
73
minimum_path_selection,
74
quotefn, sha_file, split_lines,
77
from bzrlib.testament import Testament
78
from bzrlib.trace import mutter, note, warning, is_quiet
79
from bzrlib.xml5 import serializer_v5
80
from bzrlib.inventory import InventoryEntry, make_entry
81
from bzrlib import symbol_versioning
82
from bzrlib.symbol_versioning import (deprecated_passed,
85
from bzrlib.workingtree import WorkingTree
86
from bzrlib.urlutils import unescape_for_display
90
class NullCommitReporter(object):
91
"""I report on progress of a commit."""
93
def started(self, revno, revid, location=None):
95
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
96
"to started.", DeprecationWarning,
100
def snapshot_change(self, change, path):
103
def completed(self, revno, rev_id):
106
def deleted(self, file_id):
109
def escaped(self, escape_count, message):
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 change == 'unchanged':
134
if change == 'added' and path == '':
136
self._note("%s %s", change, path)
138
def started(self, revno, rev_id, location=None):
139
if location is not None:
140
location = ' to: ' + unescape_for_display(location, 'utf-8')
142
# When started was added, location was only made optional by
143
# accident. Matt Nordhoff 20071129
144
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
145
"to started.", DeprecationWarning,
148
self._note('Committing%s', location)
150
def completed(self, revno, rev_id):
151
self._note('Committed revision %d.', revno)
153
def deleted(self, file_id):
154
self._note('deleted %s', file_id)
156
def escaped(self, escape_count, message):
157
self._note("replaced %d control characters in message", escape_count)
159
def missing(self, path):
160
self._note('missing %s', path)
162
def renamed(self, change, old_path, new_path):
163
self._note('%s %s => %s', change, old_path, new_path)
165
def is_verbose(self):
169
class Commit(object):
170
"""Task of committing a new revision.
172
This is a MethodObject: it accumulates state as the commit is
173
prepared, and then it is discarded. It doesn't represent
174
historical revisions, just the act of recording a new one.
177
Modified to hold a list of files that have been deleted from
178
the working directory; these should be removed from the
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.
184
"""Create a Commit object.
186
:param reporter: the default reporter to use or None to decide later
188
self.reporter = reporter
198
allow_pointless=True,
206
message_callback=None,
208
"""Commit working copy as a new revision.
210
:param message: the commit message (it or message_callback is required)
212
:param timestamp: if not None, seconds-since-epoch for a
213
postdated/predated commit.
215
:param specific_files: If true, commit only those files.
217
:param rev_id: If set, use this as the new revision id.
218
Useful for test or import commands that need to tightly
219
control what revisions are assigned. If you duplicate
220
a revision id that exists elsewhere it is your own fault.
221
If null (default), a time/random revision id is generated.
223
:param allow_pointless: If true (default), commit even if nothing
224
has changed and no merges are recorded.
226
:param strict: If true, don't allow a commit if the working tree
227
contains unknown files.
229
:param revprops: Properties for new revision
230
:param local: Perform a local only commit.
231
:param reporter: the reporter to use or None for the default
232
:param verbose: if True and the reporter is not None, report everything
233
:param recursive: If set to 'down', commit in any subtrees that have
234
pending changes of any sort during this commit.
236
mutter('preparing to commit')
238
if working_tree is None:
239
raise BzrError("working_tree must be passed into commit().")
241
self.work_tree = working_tree
242
self.branch = self.work_tree.branch
243
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
244
if not self.branch.repository.supports_rich_root():
245
raise errors.RootNotRich()
246
if message_callback is None:
247
if message is not None:
248
if isinstance(message, str):
249
message = message.decode(bzrlib.user_encoding)
250
message_callback = lambda x: message
252
raise BzrError("The message or message_callback keyword"
253
" parameter is required for commit().")
255
self.bound_branch = None
256
self.any_entries_changed = False
257
self.any_entries_deleted = False
259
self.master_branch = None
260
self.master_locked = False
261
self.recursive = recursive
263
if specific_files is not None:
264
self.specific_files = sorted(
265
minimum_path_selection(specific_files))
267
self.specific_files = None
268
self.specific_file_ids = None
269
self.allow_pointless = allow_pointless
270
self.revprops = revprops
271
self.message_callback = message_callback
272
self.timestamp = timestamp
273
self.timezone = timezone
274
self.committer = committer
276
self.verbose = verbose
277
# accumulates an inventory delta to the basis entry, so we can make
278
# just the necessary updates to the workingtree's cached basis.
279
self._basis_delta = []
281
self.work_tree.lock_write()
282
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
283
self.basis_revid = self.work_tree.last_revision()
284
self.basis_tree = self.work_tree.basis_tree()
285
self.basis_tree.lock_read()
287
# Cannot commit with conflicts present.
288
if len(self.work_tree.conflicts()) > 0:
289
raise ConflictsInTree
291
# Setup the bound branch variables as needed.
292
self._check_bound_branch()
294
# Check that the working tree is up to date
295
old_revno, new_revno = self._check_out_of_date_tree()
297
# Complete configuration setup
298
if reporter is not None:
299
self.reporter = reporter
300
elif self.reporter is None:
301
self.reporter = self._select_reporter()
302
if self.config is None:
303
self.config = self.branch.get_config()
305
# If provided, ensure the specified files are versioned
306
if self.specific_files is not None:
307
# Note: This routine is being called because it raises
308
# PathNotVersionedError as a side effect of finding the IDs. We
309
# later use the ids we found as input to the working tree
310
# inventory iterator, so we only consider those ids rather than
311
# examining the whole tree again.
312
# XXX: Dont we have filter_unversioned to do this more
314
self.specific_file_ids = tree.find_ids_across_trees(
315
specific_files, [self.basis_tree, self.work_tree])
317
# Setup the progress bar. As the number of files that need to be
318
# committed in unknown, progress is reported as stages.
319
# We keep track of entries separately though and include that
320
# information in the progress bar during the relevant stages.
321
self.pb_stage_name = ""
322
self.pb_stage_count = 0
323
self.pb_stage_total = 5
324
if self.bound_branch:
325
self.pb_stage_total += 1
326
self.pb.show_pct = False
327
self.pb.show_spinner = False
328
self.pb.show_eta = False
329
self.pb.show_count = True
330
self.pb.show_bar = True
332
# After a merge, a selected file commit is not supported.
333
# See 'bzr help merge' for an explanation as to why.
334
self.basis_inv = self.basis_tree.inventory
335
self._gather_parents()
336
if len(self.parents) > 1 and self.specific_files:
337
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
339
# Collect the changes
340
self._set_progress_stage("Collecting changes",
341
entries_title="Directory")
342
self.builder = self.branch.get_commit_builder(self.parents,
343
self.config, timestamp, timezone, committer, revprops, rev_id)
346
# find the location being committed to
347
if self.bound_branch:
348
master_location = self.master_branch.base
350
master_location = self.branch.base
352
# report the start of the commit
353
self.reporter.started(new_revno, self.rev_id, master_location)
355
self._update_builder_with_changes()
356
self._report_and_accumulate_deletes()
357
self._check_pointless()
359
# TODO: Now the new inventory is known, check for conflicts.
360
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
361
# weave lines, because nothing should be recorded until it is known
362
# that commit will succeed.
363
self._set_progress_stage("Saving data locally")
364
self.builder.finish_inventory()
366
# Prompt the user for a commit message if none provided
367
message = message_callback(self)
368
self.message = message
369
self._escape_commit_message()
371
# Add revision data to the local branch
372
self.rev_id = self.builder.commit(self.message)
378
self._process_pre_hooks(old_revno, new_revno)
380
# Upload revision data to the master.
381
# this will propagate merged revisions too if needed.
382
if self.bound_branch:
383
if not self.master_branch.repository.has_same_location(
384
self.branch.repository):
385
self._set_progress_stage("Uploading data to master branch")
386
self.master_branch.repository.fetch(self.branch.repository,
387
revision_id=self.rev_id)
388
# now the master has the revision data
389
# 'commit' to the master first so a timeout here causes the
390
# local branch to be out of date
391
self.master_branch.set_last_revision_info(new_revno,
394
# and now do the commit locally.
395
self.branch.set_last_revision_info(new_revno, self.rev_id)
397
# Make the working tree up to date with the branch
398
self._set_progress_stage("Updating the working tree")
399
self.work_tree.update_basis_by_delta(self.rev_id,
401
self.reporter.completed(new_revno, self.rev_id)
402
self._process_post_hooks(old_revno, new_revno)
407
def _select_reporter(self):
408
"""Select the CommitReporter to use."""
410
return NullCommitReporter()
411
return ReportCommitToLog()
413
def _check_pointless(self):
414
if self.allow_pointless:
416
# A merge with no effect on files
417
if len(self.parents) > 1:
419
# TODO: we could simplify this by using self._basis_delta.
421
# The initial commit adds a root directory, but this in itself is not
422
# a worthwhile commit.
423
if (self.basis_revid == revision.NULL_REVISION and
424
len(self.builder.new_inventory) == 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):
425
102
raise PointlessCommit()
426
# If length == 1, then we only have the root entry. Which means
427
# that there is no real difference (only the root could be different)
428
# unless deletes occured, in which case the length is irrelevant.
429
if (self.any_entries_deleted or
430
(len(self.builder.new_inventory) != 1 and
431
self.any_entries_changed)):
433
raise PointlessCommit()
435
def _check_bound_branch(self):
436
"""Check to see if the local branch is bound.
438
If it is bound, then most of the commit will actually be
439
done using the remote branch as the target branch.
440
Only at the end will the local branch be updated.
442
if self.local and not self.branch.get_bound_location():
443
raise errors.LocalRequiresBoundBranch()
446
self.master_branch = self.branch.get_master_branch()
448
if not self.master_branch:
449
# make this branch the reference branch for out of date checks.
450
self.master_branch = self.branch
453
# If the master branch is bound, we must fail
454
master_bound_location = self.master_branch.get_bound_location()
455
if master_bound_location:
456
raise errors.CommitToDoubleBoundBranch(self.branch,
457
self.master_branch, master_bound_location)
459
# TODO: jam 20051230 We could automatically push local
460
# commits to the remote branch if they would fit.
461
# But for now, just require remote to be identical
464
# Make sure the local branch is identical to the master
465
master_info = self.master_branch.last_revision_info()
466
local_info = self.branch.last_revision_info()
467
if local_info != master_info:
468
raise errors.BoundBranchOutOfDate(self.branch,
471
# Now things are ready to change the master branch
473
self.bound_branch = self.branch
474
self.master_branch.lock_write()
475
self.master_locked = True
477
def _check_out_of_date_tree(self):
478
"""Check that the working tree is up to date.
480
:return: old_revision_number,new_revision_number tuple
483
first_tree_parent = self.work_tree.get_parent_ids()[0]
485
# if there are no parents, treat our parent as 'None'
486
# this is so that we still consider the master branch
487
# - in a checkout scenario the tree may have no
488
# parents but the branch may do.
489
first_tree_parent = bzrlib.revision.NULL_REVISION
490
old_revno, master_last = self.master_branch.last_revision_info()
491
if master_last != first_tree_parent:
492
if master_last != bzrlib.revision.NULL_REVISION:
493
raise errors.OutOfDateTree(self.work_tree)
494
if self.branch.repository.has_revision(first_tree_parent):
495
new_revno = old_revno + 1
497
# ghost parents never appear in revision history.
499
return old_revno,new_revno
501
def _process_pre_hooks(self, old_revno, new_revno):
502
"""Process any registered pre commit hooks."""
503
self._set_progress_stage("Running pre_commit hooks")
504
self._process_hooks("pre_commit", old_revno, new_revno)
506
def _process_post_hooks(self, old_revno, new_revno):
507
"""Process any registered post commit hooks."""
508
# Process the post commit hooks, if any
509
self._set_progress_stage("Running post_commit hooks")
510
# old style commit hooks - should be deprecated ? (obsoleted in
512
if self.config.post_commit() is not None:
513
hooks = self.config.post_commit().split(' ')
514
# this would be nicer with twisted.python.reflect.namedAny
516
result = eval(hook + '(branch, rev_id)',
517
{'branch':self.branch,
519
'rev_id':self.rev_id})
520
# process new style post commit hooks
521
self._process_hooks("post_commit", old_revno, new_revno)
523
def _process_hooks(self, hook_name, old_revno, new_revno):
524
if not Branch.hooks[hook_name]:
527
# new style commit hooks:
528
if not self.bound_branch:
529
hook_master = self.branch
532
hook_master = self.master_branch
533
hook_local = self.branch
534
# With bound branches, when the master is behind the local branch,
535
# the 'old_revno' and old_revid values here are incorrect.
536
# XXX: FIXME ^. RBC 20060206
538
old_revid = self.parents[0]
540
old_revid = bzrlib.revision.NULL_REVISION
542
if hook_name == "pre_commit":
543
future_tree = self.builder.revision_tree()
544
tree_delta = future_tree.changes_from(self.basis_tree,
547
for hook in Branch.hooks[hook_name]:
548
# show the running hook in the progress bar. As hooks may
549
# end up doing nothing (e.g. because they are not configured by
550
# the user) this is still showing progress, not showing overall
551
# actions - its up to each plugin to show a UI if it want's to
552
# (such as 'Emailing diff to foo@example.com').
553
self.pb_stage_name = "Running %s hooks [%s]" % \
554
(hook_name, Branch.hooks.get_hook_name(hook))
555
self._emit_progress()
556
if 'hooks' in debug.debug_flags:
557
mutter("Invoking commit hook: %r", hook)
558
if hook_name == "post_commit":
559
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
561
elif hook_name == "pre_commit":
562
hook(hook_local, hook_master,
563
old_revno, old_revid, new_revno, self.rev_id,
564
tree_delta, future_tree)
567
"""Cleanup any open locks, progress bars etc."""
568
cleanups = [self._cleanup_bound_branch,
569
self.basis_tree.unlock,
570
self.work_tree.unlock,
572
found_exception = None
573
for cleanup in cleanups:
576
# we want every cleanup to run no matter what.
577
# so we have a catchall here, but we will raise the
578
# last encountered exception up the stack: and
579
# typically this will be useful enough.
582
if found_exception is not None:
583
# don't do a plan raise, because the last exception may have been
584
# trashed, e is our sure-to-work exception even though it loses the
585
# full traceback. XXX: RBC 20060421 perhaps we could check the
586
# exc_info and if its the same one do a plain raise otherwise
587
# 'raise e' as we do now.
590
def _cleanup_bound_branch(self):
591
"""Executed at the end of a try/finally to cleanup a bound branch.
593
If the branch wasn't bound, this is a no-op.
594
If it was, it resents self.branch to the local branch, instead
597
if not self.bound_branch:
599
if self.master_locked:
600
self.master_branch.unlock()
602
def _escape_commit_message(self):
603
"""Replace xml-incompatible control characters."""
604
# FIXME: RBC 20060419 this should be done by the revision
605
# serialiser not by commit. Then we can also add an unescaper
606
# in the deserializer and start roundtripping revision messages
607
# precisely. See repository_implementations/test_repository.py
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")
609
147
# Python strings can include characters that can't be
610
148
# represented in well-formed XML; escape characters that
611
149
# aren't listed in the XML specification
612
150
# (http://www.w3.org/TR/REC-xml/#NT-Char).
613
self.message, escape_count = re.subn(
614
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
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(
615
159
lambda match: match.group(0).encode('unicode_escape'),
618
self.reporter.escaped(escape_count, self.message)
620
def _gather_parents(self):
621
"""Record the parents of a merge for merge detection."""
622
# TODO: Make sure that this list doesn't contain duplicate
623
# entries and the order is preserved when doing this.
624
self.parents = self.work_tree.get_parent_ids()
625
self.parent_invs = [self.basis_inv]
626
for revision in self.parents[1:]:
627
if self.branch.repository.has_revision(revision):
628
mutter('commit parent revision {%s}', revision)
629
inventory = self.branch.repository.get_inventory(revision)
630
self.parent_invs.append(inventory)
632
mutter('commit parent ghost revision {%s}', revision)
634
def _update_builder_with_changes(self):
635
"""Update the commit builder with the data about what has changed.
637
# Build the revision inventory.
639
# This starts by creating a new empty inventory. Depending on
640
# which files are selected for commit, and what is present in the
641
# current tree, the new inventory is populated. inventory entries
642
# which are candidates for modification have their revision set to
643
# None; inventory entries that are carried over untouched have their
644
# revision set to their prior value.
646
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
647
# results to create a new inventory at the same time, which results
648
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
651
specific_files = self.specific_files
652
mutter("Selecting files for commit with filter %s", specific_files)
654
# Build the new inventory
655
self._populate_from_inventory(specific_files)
657
# If specific files are selected, then all un-selected files must be
658
# recorded in their previous state. For more details, see
659
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
661
for path, old_ie in self.basis_inv.iter_entries():
662
if old_ie.file_id in self.builder.new_inventory:
663
# already added - skip.
665
if is_inside_any(specific_files, path):
666
# was inside the selected path, if not present it has been
669
if old_ie.kind == 'directory':
670
self._next_progress_entry()
671
# not in final inv yet, was not in the selected files, so is an
672
# entry to be preserved unaltered.
674
# Note: specific file commits after a merge are currently
675
# prohibited. This test is for sanity/safety in case it's
676
# required after that changes.
677
if len(self.parents) > 1:
679
delta, version_recorded = self.builder.record_entry_contents(
680
ie, self.parent_invs, path, self.basis_tree, None)
682
self.any_entries_changed = True
683
if delta: self._basis_delta.append(delta)
685
def _report_and_accumulate_deletes(self):
686
# XXX: Could the list of deleted paths and ids be instead taken from
687
# _populate_from_inventory?
688
deleted_ids = set(self.basis_inv._byid.keys()) - \
689
set(self.builder.new_inventory._byid.keys())
691
self.any_entries_deleted = True
692
deleted = [(self.basis_tree.id2path(file_id), file_id)
693
for file_id in deleted_ids]
695
# XXX: this is not quite directory-order sorting
696
for path, file_id in deleted:
697
self._basis_delta.append((path, None, file_id, None))
698
self.reporter.deleted(path)
700
def _populate_from_inventory(self, specific_files):
701
"""Populate the CommitBuilder by walking the working tree inventory."""
703
# raise an exception as soon as we find a single unknown.
704
for unknown in self.work_tree.unknowns():
705
raise StrictCommitFailed()
707
report_changes = self.reporter.is_verbose()
709
# A tree of paths that have been deleted. E.g. if foo/bar has been
710
# deleted, then we have {'foo':{'bar':{}}}
712
# XXX: Note that entries may have the wrong kind because the entry does
713
# not reflect the status on disk.
714
work_inv = self.work_tree.inventory
715
entries = work_inv.iter_entries_by_dir(
716
specific_file_ids=self.specific_file_ids, yield_parents=True)
717
for path, existing_ie in entries:
718
file_id = existing_ie.file_id
719
name = existing_ie.name
720
parent_id = existing_ie.parent_id
721
kind = existing_ie.kind
722
if kind == 'directory':
723
self._next_progress_entry()
724
# Skip files that have been deleted from the working tree.
725
# The deleted path ids are also recorded so they can be explicitly
728
path_segments = splitpath(path)
729
deleted_dict = deleted_paths
730
for segment in path_segments:
731
deleted_dict = deleted_dict.get(segment, None)
733
# We either took a path not present in the dict
734
# (deleted_dict was None), or we've reached an empty
735
# child dir in the dict, so are now a sub-path.
739
if deleted_dict is not None:
740
# the path has a deleted parent, do not add it.
742
content_summary = self.work_tree.path_content_summary(path)
743
# Note that when a filter of specific files is given, we must only
744
# skip/record deleted files matching that filter.
745
if not specific_files or is_inside_any(specific_files, path):
746
if content_summary[0] == 'missing':
747
if not deleted_paths:
748
# path won't have been split yet.
749
path_segments = splitpath(path)
750
deleted_dict = deleted_paths
751
for segment in path_segments:
752
deleted_dict = deleted_dict.setdefault(segment, {})
753
self.reporter.missing(path)
754
deleted_ids.append(file_id)
756
# TODO: have the builder do the nested commit just-in-time IF and
758
if content_summary[0] == 'tree-reference':
759
# enforce repository nested tree policy.
760
if (not self.work_tree.supports_tree_reference() or
761
# repository does not support it either.
762
not self.branch.repository._format.supports_tree_reference):
763
content_summary = ('directory',) + content_summary[1:]
764
kind = content_summary[0]
765
# TODO: specific_files filtering before nested tree processing
766
if kind == 'tree-reference':
767
if self.recursive == 'down':
768
nested_revision_id = self._commit_nested_tree(
770
content_summary = content_summary[:3] + (
773
content_summary = content_summary[:3] + (
774
self.work_tree.get_reference_revision(file_id),)
776
# Record an entry for this item
777
# Note: I don't particularly want to have the existing_ie
778
# parameter but the test suite currently (28-Jun-07) breaks
779
# without it thanks to a unicode normalisation issue. :-(
780
definitely_changed = kind != existing_ie.kind
781
self._record_entry(path, file_id, specific_files, kind, name,
782
parent_id, definitely_changed, existing_ie, report_changes,
785
# Unversion IDs that were found to be deleted
786
self.work_tree.unversion(deleted_ids)
788
def _commit_nested_tree(self, file_id, path):
789
"Commit a nested tree."
790
sub_tree = self.work_tree.get_nested_tree(file_id, path)
791
# FIXME: be more comprehensive here:
792
# this works when both trees are in --trees repository,
793
# but when both are bound to a different repository,
794
# it fails; a better way of approaching this is to
795
# finally implement the explicit-caches approach design
796
# a while back - RBC 20070306.
797
if sub_tree.branch.repository.has_same_location(
798
self.work_tree.branch.repository):
799
sub_tree.branch.repository = \
800
self.work_tree.branch.repository
802
return sub_tree.commit(message=None, revprops=self.revprops,
803
recursive=self.recursive,
804
message_callback=self.message_callback,
805
timestamp=self.timestamp, timezone=self.timezone,
806
committer=self.committer,
807
allow_pointless=self.allow_pointless,
808
strict=self.strict, verbose=self.verbose,
809
local=self.local, reporter=self.reporter)
810
except errors.PointlessCommit:
811
return self.work_tree.get_reference_revision(file_id)
813
def _record_entry(self, path, file_id, specific_files, kind, name,
814
parent_id, definitely_changed, existing_ie, report_changes,
816
"Record the new inventory entry for a path if any."
817
# mutter('check %s {%s}', path, file_id)
818
# mutter('%s selected for commit', path)
819
if definitely_changed or existing_ie is None:
820
ie = make_entry(kind, name, parent_id, file_id)
822
ie = existing_ie.copy()
824
delta, version_recorded = self.builder.record_entry_contents(ie,
825
self.parent_invs, path, self.work_tree, content_summary)
827
self._basis_delta.append(delta)
829
self.any_entries_changed = True
831
self._report_change(ie, path)
834
def _report_change(self, ie, path):
835
"""Report a change to the user.
837
The change that has occurred is described relative to the basis
840
if (self.basis_inv.has_id(ie.file_id)):
841
basis_ie = self.basis_inv[ie.file_id]
844
change = ie.describe_change(basis_ie, ie)
845
if change in (InventoryEntry.RENAMED,
846
InventoryEntry.MODIFIED_AND_RENAMED):
847
old_path = self.basis_inv.id2path(ie.file_id)
848
self.reporter.renamed(change, old_path, path)
850
self.reporter.snapshot_change(change, path)
852
def _set_progress_stage(self, name, entries_title=None):
853
"""Set the progress stage and emit an update to the progress bar."""
854
self.pb_stage_name = name
855
self.pb_stage_count += 1
856
self.pb_entries_title = entries_title
857
if entries_title is not None:
858
self.pb_entries_count = 0
859
self.pb_entries_total = '?'
860
self._emit_progress()
862
def _next_progress_entry(self):
863
"""Emit an update to the progress bar and increment the entry count."""
864
self.pb_entries_count += 1
865
self._emit_progress()
867
def _emit_progress(self):
868
if self.pb_entries_title:
869
if self.pb_entries_total == '?':
870
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
871
self.pb_entries_title, self.pb_entries_count)
873
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
874
self.pb_entries_title, self.pb_entries_count,
875
str(self.pb_entries_total))
877
text = "%s - Stage" % (self.pb_stage_name)
878
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
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
