1
# Copyright (C) 2005, 2006 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
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 inventory. 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 inventory, 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
64
from bzrlib.branch import Branch
66
from bzrlib.errors import (BzrError, PointlessCommit,
70
from bzrlib.osutils import (kind_marker, isdir,isfile, is_inside_any,
71
is_inside_or_parent_of_any,
72
quotefn, sha_file, split_lines)
73
from bzrlib.testament import Testament
74
from bzrlib.trace import mutter, note, warning
75
from bzrlib.xml5 import serializer_v5
76
from bzrlib.inventory import Inventory, InventoryEntry
77
from bzrlib import symbol_versioning
78
from bzrlib.symbol_versioning import (deprecated_passed,
81
from bzrlib.workingtree import WorkingTree
85
class NullCommitReporter(object):
86
"""I report on progress of a commit."""
88
def snapshot_change(self, change, path):
91
def completed(self, revno, rev_id):
94
def deleted(self, file_id):
97
def escaped(self, escape_count, message):
100
def missing(self, path):
103
def renamed(self, change, old_path, new_path):
107
class ReportCommitToLog(NullCommitReporter):
109
# this may be more useful if 'note' was replaced by an overridable
110
# method on self, which would allow more trivial subclassing.
111
# alternative, a callable could be passed in, allowing really trivial
112
# reuse for some uis. RBC 20060511
114
def snapshot_change(self, change, path):
115
if change == 'unchanged':
117
if change == 'added' and path == '':
119
note("%s %s", change, path)
121
def completed(self, revno, rev_id):
122
note('Committed revision %d.', revno)
124
def deleted(self, file_id):
125
note('deleted %s', file_id)
127
def escaped(self, escape_count, message):
128
note("replaced %d control characters in message", escape_count)
130
def missing(self, path):
131
note('missing %s', path)
133
def renamed(self, change, old_path, new_path):
134
note('%s %s => %s', change, old_path, new_path)
137
class Commit(object):
138
"""Task of committing a new revision.
140
This is a MethodObject: it accumulates state as the commit is
141
prepared, and then it is discarded. It doesn't represent
142
historical revisions, just the act of recording a new one.
145
Modified to hold a list of files that have been deleted from
146
the working directory; these should be removed from the
152
if reporter is not None:
153
self.reporter = reporter
155
self.reporter = NullCommitReporter()
156
if config is not None:
162
branch=DEPRECATED_PARAMETER, message=None,
168
allow_pointless=True,
176
message_callback=None,
178
"""Commit working copy as a new revision.
180
branch -- the deprecated branch to commit to. New callers should pass in
183
message -- the commit message (it or message_callback is required)
185
timestamp -- if not None, seconds-since-epoch for a
186
postdated/predated commit.
188
specific_files -- If true, commit only those files.
190
rev_id -- If set, use this as the new revision id.
191
Useful for test or import commands that need to tightly
192
control what revisions are assigned. If you duplicate
193
a revision id that exists elsewhere it is your own fault.
194
If null (default), a time/random revision id is generated.
196
allow_pointless -- If true (default), commit even if nothing
197
has changed and no merges are recorded.
199
strict -- If true, don't allow a commit if the working tree
200
contains unknown files.
202
revprops -- Properties for new revision
203
:param local: Perform a local only commit.
204
:param recursive: If set to 'down', commit in any subtrees that have
205
pending changes of any sort during this commit.
207
mutter('preparing to commit')
209
if deprecated_passed(branch):
210
symbol_versioning.warn("Commit.commit (branch, ...): The branch parameter is "
211
"deprecated as of bzr 0.8. Please use working_tree= instead.",
212
DeprecationWarning, stacklevel=2)
214
self.work_tree = self.branch.bzrdir.open_workingtree()
215
elif working_tree is None:
216
raise BzrError("One of branch and working_tree must be passed into commit().")
218
self.work_tree = working_tree
219
self.branch = self.work_tree.branch
220
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
221
if not self.branch.repository.supports_rich_root():
222
raise errors.RootNotRich()
223
if message_callback is None:
224
if message is not None:
225
if isinstance(message, str):
226
message = message.decode(bzrlib.user_encoding)
227
message_callback = lambda x: message
229
raise BzrError("The message or message_callback keyword"
230
" parameter is required for commit().")
232
self.bound_branch = None
234
self.master_branch = None
235
self.master_locked = False
237
self.specific_files = specific_files
238
self.allow_pointless = allow_pointless
239
self.recursive = recursive
240
self.revprops = revprops
241
self.message_callback = message_callback
242
self.timestamp = timestamp
243
self.timezone = timezone
244
self.committer = committer
245
self.specific_files = specific_files
247
self.verbose = verbose
250
if reporter is None and self.reporter is None:
251
self.reporter = NullCommitReporter()
252
elif reporter is not None:
253
self.reporter = reporter
255
self.work_tree.lock_write()
256
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
257
self.basis_tree = self.work_tree.basis_tree()
258
self.basis_tree.lock_read()
260
# Cannot commit with conflicts present.
261
if len(self.work_tree.conflicts())>0:
262
raise ConflictsInTree
264
# setup the bound branch variables as needed.
265
self._check_bound_branch()
267
# check for out of date working trees
269
first_tree_parent = self.work_tree.get_parent_ids()[0]
271
# if there are no parents, treat our parent as 'None'
272
# this is so that we still consier the master branch
273
# - in a checkout scenario the tree may have no
274
# parents but the branch may do.
275
first_tree_parent = bzrlib.revision.NULL_REVISION
276
old_revno, master_last = self.master_branch.last_revision_info()
277
if master_last != first_tree_parent:
278
if master_last != bzrlib.revision.NULL_REVISION:
279
raise errors.OutOfDateTree(self.work_tree)
280
if self.branch.repository.has_revision(first_tree_parent):
281
new_revno = old_revno + 1
283
# ghost parents never appear in revision history.
286
# raise an exception as soon as we find a single unknown.
287
for unknown in self.work_tree.unknowns():
288
raise StrictCommitFailed()
290
if self.config is None:
291
self.config = self.branch.get_config()
293
self.work_inv = self.work_tree.inventory
294
self.basis_inv = self.basis_tree.inventory
295
if specific_files is not None:
296
# Ensure specified files are versioned
297
# (We don't actually need the ids here)
298
# XXX: Dont we have filter_unversioned to do this more
300
tree.find_ids_across_trees(specific_files,
301
[self.basis_tree, self.work_tree])
302
# one to finish, one for rev and inventory, and one for each
303
# inventory entry, and the same for the new inventory.
304
# note that this estimate is too long when we do a partial tree
305
# commit which excludes some new files from being considered.
306
# The estimate is corrected when we populate the new inv.
307
self.pb_total = len(self.work_inv) + 5
310
self._gather_parents()
311
if len(self.parents) > 1 and self.specific_files:
312
raise NotImplementedError('selected-file commit of merges is not supported yet: files %r',
315
self.builder = self.branch.get_commit_builder(self.parents,
316
self.config, timestamp, timezone, committer, revprops, rev_id)
318
self._remove_deleted()
319
self._populate_new_inv()
320
self._report_deletes()
322
self._check_pointless()
324
self._emit_progress_update()
325
# TODO: Now the new inventory is known, check for conflicts and
326
# prompt the user for a commit message.
327
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
328
# weave lines, because nothing should be recorded until it is known
329
# that commit will succeed.
330
self.builder.finish_inventory()
331
self._emit_progress_update()
332
message = message_callback(self)
333
assert isinstance(message, unicode), type(message)
334
self.message = message
335
self._escape_commit_message()
337
self.rev_id = self.builder.commit(self.message)
338
self._emit_progress_update()
339
# revision data is in the local branch now.
341
# upload revision data to the master.
342
# this will propagate merged revisions too if needed.
343
if self.bound_branch:
344
self.master_branch.repository.fetch(self.branch.repository,
345
revision_id=self.rev_id)
346
# now the master has the revision data
347
# 'commit' to the master first so a timeout here causes the local
348
# branch to be out of date
349
self.master_branch.set_last_revision_info(new_revno,
352
# and now do the commit locally.
353
self.branch.set_last_revision_info(new_revno, self.rev_id)
355
rev_tree = self.builder.revision_tree()
356
self.work_tree.set_parent_trees([(self.rev_id, rev_tree)])
357
# now the work tree is up to date with the branch
359
self.reporter.completed(new_revno, self.rev_id)
360
# old style commit hooks - should be deprecated ? (obsoleted in
362
if self.config.post_commit() is not None:
363
hooks = self.config.post_commit().split(' ')
364
# this would be nicer with twisted.python.reflect.namedAny
366
result = eval(hook + '(branch, rev_id)',
367
{'branch':self.branch,
369
'rev_id':self.rev_id})
370
# new style commit hooks:
371
if not self.bound_branch:
372
hook_master = self.branch
375
hook_master = self.master_branch
376
hook_local = self.branch
377
# With bound branches, when the master is behind the local branch,
378
# the 'old_revno' and old_revid values here are incorrect.
379
# XXX: FIXME ^. RBC 20060206
381
old_revid = self.parents[0]
383
old_revid = bzrlib.revision.NULL_REVISION
384
for hook in Branch.hooks['post_commit']:
385
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
387
self._emit_progress_update()
392
def _any_real_changes(self):
393
"""Are there real changes between new_inventory and basis?
395
For trees without rich roots, inv.root.revision changes every commit.
396
But if that is the only change, we want to treat it as though there
399
new_entries = self.builder.new_inventory.iter_entries()
400
basis_entries = self.basis_inv.iter_entries()
401
new_path, new_root_ie = new_entries.next()
402
basis_path, basis_root_ie = basis_entries.next()
404
# This is a copy of InventoryEntry.__eq__ only leaving out .revision
405
def ie_equal_no_revision(this, other):
406
return ((this.file_id == other.file_id)
407
and (this.name == other.name)
408
and (this.symlink_target == other.symlink_target)
409
and (this.text_sha1 == other.text_sha1)
410
and (this.text_size == other.text_size)
411
and (this.text_id == other.text_id)
412
and (this.parent_id == other.parent_id)
413
and (this.kind == other.kind)
414
and (this.executable == other.executable)
415
and (this.reference_revision == other.reference_revision)
417
if not ie_equal_no_revision(new_root_ie, basis_root_ie):
420
for new_ie, basis_ie in zip(new_entries, basis_entries):
421
if new_ie != basis_ie:
424
# No actual changes present
427
def _check_pointless(self):
428
if self.allow_pointless:
430
# A merge with no effect on files
431
if len(self.parents) > 1:
433
# work around the fact that a newly-initted tree does differ from its
435
if len(self.basis_inv) == 0 and len(self.builder.new_inventory) == 1:
436
raise PointlessCommit()
437
# Shortcut, if the number of entries changes, then we obviously have
439
if len(self.builder.new_inventory) != len(self.basis_inv):
441
# If length == 1, then we only have the root entry. Which means
442
# that there is no real difference (only the root could be different)
443
if (len(self.builder.new_inventory) != 1 and self._any_real_changes()):
445
raise PointlessCommit()
447
def _check_bound_branch(self):
448
"""Check to see if the local branch is bound.
450
If it is bound, then most of the commit will actually be
451
done using the remote branch as the target branch.
452
Only at the end will the local branch be updated.
454
if self.local and not self.branch.get_bound_location():
455
raise errors.LocalRequiresBoundBranch()
458
self.master_branch = self.branch.get_master_branch()
460
if not self.master_branch:
461
# make this branch the reference branch for out of date checks.
462
self.master_branch = self.branch
465
# If the master branch is bound, we must fail
466
master_bound_location = self.master_branch.get_bound_location()
467
if master_bound_location:
468
raise errors.CommitToDoubleBoundBranch(self.branch,
469
self.master_branch, master_bound_location)
471
# TODO: jam 20051230 We could automatically push local
472
# commits to the remote branch if they would fit.
473
# But for now, just require remote to be identical
476
# Make sure the local branch is identical to the master
477
master_info = self.master_branch.last_revision_info()
478
local_info = self.branch.last_revision_info()
479
if local_info != master_info:
480
raise errors.BoundBranchOutOfDate(self.branch,
483
# Now things are ready to change the master branch
485
self.bound_branch = self.branch
486
self.master_branch.lock_write()
487
self.master_locked = True
490
"""Cleanup any open locks, progress bars etc."""
491
cleanups = [self._cleanup_bound_branch,
492
self.basis_tree.unlock,
493
self.work_tree.unlock,
495
found_exception = None
496
for cleanup in cleanups:
499
# we want every cleanup to run no matter what.
500
# so we have a catchall here, but we will raise the
501
# last encountered exception up the stack: and
502
# typically this will be useful enough.
505
if found_exception is not None:
506
# don't do a plan raise, because the last exception may have been
507
# trashed, e is our sure-to-work exception even though it loses the
508
# full traceback. XXX: RBC 20060421 perhaps we could check the
509
# exc_info and if its the same one do a plain raise otherwise
510
# 'raise e' as we do now.
513
def _cleanup_bound_branch(self):
514
"""Executed at the end of a try/finally to cleanup a bound branch.
516
If the branch wasn't bound, this is a no-op.
517
If it was, it resents self.branch to the local branch, instead
520
if not self.bound_branch:
522
if self.master_locked:
523
self.master_branch.unlock()
525
def _escape_commit_message(self):
526
"""Replace xml-incompatible control characters."""
527
# FIXME: RBC 20060419 this should be done by the revision
528
# serialiser not by commit. Then we can also add an unescaper
529
# in the deserializer and start roundtripping revision messages
530
# precisely. See repository_implementations/test_repository.py
532
# Python strings can include characters that can't be
533
# represented in well-formed XML; escape characters that
534
# aren't listed in the XML specification
535
# (http://www.w3.org/TR/REC-xml/#NT-Char).
536
self.message, escape_count = re.subn(
537
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
538
lambda match: match.group(0).encode('unicode_escape'),
541
self.reporter.escaped(escape_count, self.message)
543
def _gather_parents(self):
544
"""Record the parents of a merge for merge detection."""
545
# TODO: Make sure that this list doesn't contain duplicate
546
# entries and the order is preserved when doing this.
547
self.parents = self.work_tree.get_parent_ids()
548
self.parent_invs = [self.basis_inv]
549
for revision in self.parents[1:]:
550
if self.branch.repository.has_revision(revision):
551
mutter('commit parent revision {%s}', revision)
552
inventory = self.branch.repository.get_inventory(revision)
553
self.parent_invs.append(inventory)
555
mutter('commit parent ghost revision {%s}', revision)
557
def _remove_deleted(self):
558
"""Remove deleted files from the working inventories.
560
This is done prior to taking the working inventory as the
561
basis for the new committed inventory.
563
This returns true if any files
564
*that existed in the basis inventory* were deleted.
565
Files that were added and deleted
566
in the working copy don't matter.
568
specific = self.specific_files
570
deleted_paths = set()
571
for path, ie in self.work_inv.iter_entries():
572
if is_inside_any(deleted_paths, path):
573
# The tree will delete the required ids recursively.
575
if specific and not is_inside_any(specific, path):
577
if not self.work_tree.has_filename(path):
578
deleted_paths.add(path)
579
self.reporter.missing(path)
580
deleted_ids.append(ie.file_id)
581
self.work_tree.unversion(deleted_ids)
583
def _populate_new_inv(self):
584
"""Build revision inventory.
586
This creates a new empty inventory. Depending on
587
which files are selected for commit, and what is present in the
588
current tree, the new inventory is populated. inventory entries
589
which are candidates for modification have their revision set to
590
None; inventory entries that are carried over untouched have their
591
revision set to their prior value.
593
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
594
# results to create a new inventory at the same time, which results
595
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
597
mutter("Selecting files for commit with filter %s", self.specific_files)
598
assert self.work_inv.root is not None
599
entries = self.work_inv.iter_entries()
600
if not self.builder.record_root_entry:
601
symbol_versioning.warn('CommitBuilders should support recording'
602
' the root entry as of bzr 0.10.', DeprecationWarning,
604
self.builder.new_inventory.add(self.basis_inv.root.copy())
606
self._emit_progress_update()
607
for path, new_ie in entries:
608
self._emit_progress_update()
609
file_id = new_ie.file_id
611
kind = self.work_tree.kind(file_id)
612
if kind == 'tree-reference' and self.recursive == 'down':
613
# nested tree: commit in it
614
sub_tree = WorkingTree.open(self.work_tree.abspath(path))
615
# FIXME: be more comprehensive here:
616
# this works when both trees are in --trees repository,
617
# but when both are bound to a different repository,
618
# it fails; a better way of approaching this is to
619
# finally implement the explicit-caches approach design
620
# a while back - RBC 20070306.
621
if (sub_tree.branch.repository.bzrdir.root_transport.base
623
self.work_tree.branch.repository.bzrdir.root_transport.base):
624
sub_tree.branch.repository = \
625
self.work_tree.branch.repository
627
sub_tree.commit(message=None, revprops=self.revprops,
628
recursive=self.recursive,
629
message_callback=self.message_callback,
630
timestamp=self.timestamp, timezone=self.timezone,
631
committer=self.committer,
632
allow_pointless=self.allow_pointless,
633
strict=self.strict, verbose=self.verbose,
634
local=self.local, reporter=self.reporter)
635
except errors.PointlessCommit:
637
if kind != new_ie.kind:
638
new_ie = inventory.make_entry(kind, new_ie.name,
639
new_ie.parent_id, file_id)
640
except errors.NoSuchFile:
642
# mutter('check %s {%s}', path, file_id)
643
if (not self.specific_files or
644
is_inside_or_parent_of_any(self.specific_files, path)):
645
# mutter('%s selected for commit', path)
649
# mutter('%s not selected for commit', path)
650
if self.basis_inv.has_id(file_id):
651
ie = self.basis_inv[file_id].copy()
653
# this entry is new and not being committed
655
self.builder.record_entry_contents(ie, self.parent_invs,
656
path, self.work_tree)
657
# describe the nature of the change that has occurred relative to
658
# the basis inventory.
659
if (self.basis_inv.has_id(ie.file_id)):
660
basis_ie = self.basis_inv[ie.file_id]
663
change = ie.describe_change(basis_ie, ie)
664
if change in (InventoryEntry.RENAMED,
665
InventoryEntry.MODIFIED_AND_RENAMED):
666
old_path = self.basis_inv.id2path(ie.file_id)
667
self.reporter.renamed(change, old_path, path)
669
self.reporter.snapshot_change(change, path)
671
if not self.specific_files:
674
# ignore removals that don't match filespec
675
for path, new_ie in self.basis_inv.iter_entries():
676
if new_ie.file_id in self.work_inv:
678
if is_inside_any(self.specific_files, path):
682
self.builder.record_entry_contents(ie, self.parent_invs, path,
685
def _emit_progress_update(self):
686
"""Emit an update to the progress bar."""
687
self.pb.update("Committing", self.pb_count, self.pb_total)
690
def _report_deletes(self):
691
for path, ie in self.basis_inv.iter_entries():
692
if ie.file_id not in self.builder.new_inventory:
693
self.reporter.deleted(path)