1
# Copyright (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007 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
19
def commit(branch, message, timestamp=None, timezone=None,
22
"""Commit working copy as a new revision.
24
The basic approach is to add all the file texts into the
25
store, then the inventory, then make a new revision pointing
26
to that inventory and store that.
28
This is not quite safe if the working copy changes during the
29
commit; for the moment that is simply not allowed. A better
30
approach is to make a temporary copy of the files before
31
computing their hashes, and then add those hashes in turn to
32
the inventory. This should mean at least that there are no
33
broken hash pointers. There is no way we can get a snapshot
34
of the whole directory at an instant. This would also have to
35
be robust against files disappearing, moving, etc. So the
36
whole thing is a bit hard.
38
timestamp -- if not None, seconds-since-epoch for a
39
postdated/predated commit.
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 (get_user_encoding,
72
kind_marker, isdir,isfile, is_inside_any,
73
is_inside_or_parent_of_any,
74
minimum_path_selection,
75
quotefn, sha_file, split_lines,
78
from bzrlib.testament import Testament
79
from bzrlib.trace import mutter, note, warning, is_quiet
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
42
import os, time, tempfile
44
from inventory import Inventory
45
from osutils import isdir, isfile, sha_string, quotefn, \
46
local_time_offset, username
48
from branch import gen_file_id
49
from errors import BzrError
50
from revision import Revision
51
from textui import show_status
52
from trace import mutter, note
54
branch._need_writelock()
56
## TODO: Show branch names
58
# TODO: Don't commit if there are no changes, unless forced?
60
# First walk over the working inventory; and both update that
61
# and also build a new revision inventory. The revision
62
# inventory needs to hold the text-id, sha1 and size of the
63
# actual file versions committed in the revision. (These are
64
# not present in the working inventory.) We also need to
65
# detect missing/deleted files, and remove them from the
68
work_inv = branch.read_working_inventory()
70
basis = branch.basis_tree()
71
basis_inv = basis.inventory
73
for path, entry in work_inv.iter_entries():
74
## TODO: Cope with files that have gone missing.
76
## TODO: Check that the file kind has not changed from the previous
77
## revision of this file (if any).
81
p = branch.abspath(path)
82
file_id = entry.file_id
83
mutter('commit prep file %s, id %r ' % (p, file_id))
85
if not os.path.exists(p):
86
mutter(" file is missing, removing from inventory")
88
show_status('D', entry.kind, quotefn(path))
89
missing_ids.append(file_id)
92
# TODO: Handle files that have been deleted
94
# TODO: Maybe a special case for empty files? Seems a
95
# waste to store them many times.
99
if basis_inv.has_id(file_id):
100
old_kind = basis_inv[file_id].kind
101
if old_kind != entry.kind:
102
raise BzrError("entry %r changed kind from %r to %r"
103
% (file_id, old_kind, entry.kind))
105
if entry.kind == 'directory':
107
raise BzrError("%s is entered as directory but not a directory" % quotefn(p))
108
elif entry.kind == 'file':
110
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
112
content = file(p, 'rb').read()
114
entry.text_sha1 = sha_string(content)
115
entry.text_size = len(content)
117
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
119
and (old_ie.text_size == entry.text_size)
120
and (old_ie.text_sha1 == entry.text_sha1)):
121
## assert content == basis.get_file(file_id).read()
122
entry.text_id = basis_inv[file_id].text_id
123
mutter(' unchanged from previous text_id {%s}' %
127
entry.text_id = gen_file_id(entry.name)
128
branch.text_store.add(content, entry.text_id)
129
mutter(' stored with text_id {%s}' % entry.text_id)
133
elif (old_ie.name == entry.name
134
and old_ie.parent_id == entry.parent_id):
139
show_status(state, entry.kind, quotefn(path))
141
for file_id in missing_ids:
142
# have to do this later so we don't mess up the iterator.
143
# since parents may be removed before their children we
146
# FIXME: There's probably a better way to do this; perhaps
147
# the workingtree should know how to filter itbranch.
148
if work_inv.has_id(file_id):
149
del work_inv[file_id]
152
inv_id = rev_id = _gen_revision_id(time.time())
154
inv_tmp = tempfile.TemporaryFile()
155
inv.write_xml(inv_tmp)
157
branch.inventory_store.add(inv_tmp, inv_id)
158
mutter('new inventory_id is {%s}' % inv_id)
160
branch._write_inventory(work_inv)
162
if timestamp == None:
163
timestamp = time.time()
165
if committer == None:
166
committer = username()
169
timezone = local_time_offset()
171
mutter("building commit log message")
172
rev = Revision(timestamp=timestamp,
175
precursor = branch.last_patch(),
180
rev_tmp = tempfile.TemporaryFile()
181
rev.write_xml(rev_tmp)
183
branch.revision_store.add(rev_tmp, rev_id)
184
mutter("new revision_id is {%s}" % rev_id)
186
## XXX: Everything up to here can simply be orphaned if we abort
187
## the commit; it will leave junk files behind but that doesn't
190
## TODO: Read back the just-generated changeset, and make sure it
191
## applies and recreates the right state.
193
## TODO: Also calculate and store the inventory SHA1
194
mutter("committing patch r%d" % (branch.revno() + 1))
197
branch.append_revision(rev_id)
200
note("commited r%d" % branch.revno())
204
def _gen_revision_id(when):
205
"""Return new revision-id."""
206
from binascii import hexlify
207
from osutils import rand_bytes, compact_date, user_email
209
s = '%s-%s-' % (user_email(), compact_date(when))
210
s += hexlify(rand_bytes(8))
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,
209
possible_master_transports=None):
210
"""Commit working copy as a new revision.
212
:param message: the commit message (it or message_callback is required)
214
:param timestamp: if not None, seconds-since-epoch for a
215
postdated/predated commit.
217
:param specific_files: If true, commit only those files.
219
:param rev_id: If set, use this as the new revision id.
220
Useful for test or import commands that need to tightly
221
control what revisions are assigned. If you duplicate
222
a revision id that exists elsewhere it is your own fault.
223
If null (default), a time/random revision id is generated.
225
:param allow_pointless: If true (default), commit even if nothing
226
has changed and no merges are recorded.
228
:param strict: If true, don't allow a commit if the working tree
229
contains unknown files.
231
:param revprops: Properties for new revision
232
:param local: Perform a local only commit.
233
:param reporter: the reporter to use or None for the default
234
:param verbose: if True and the reporter is not None, report everything
235
:param recursive: If set to 'down', commit in any subtrees that have
236
pending changes of any sort during this commit.
237
:param exclude: None or a list of relative paths to exclude from the
238
commit. Pending changes to excluded files will be ignored by the
241
mutter('preparing to commit')
243
if working_tree is None:
244
raise BzrError("working_tree must be passed into commit().")
246
self.work_tree = working_tree
247
self.branch = self.work_tree.branch
248
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
249
if not self.branch.repository.supports_rich_root():
250
raise errors.RootNotRich()
251
if message_callback is None:
252
if message is not None:
253
if isinstance(message, str):
254
message = message.decode(get_user_encoding())
255
message_callback = lambda x: message
257
raise BzrError("The message or message_callback keyword"
258
" parameter is required for commit().")
260
self.bound_branch = None
261
self.any_entries_changed = False
262
self.any_entries_deleted = False
263
if exclude is not None:
264
self.exclude = sorted(
265
minimum_path_selection(exclude))
269
self.master_branch = None
270
self.master_locked = False
271
self.recursive = recursive
273
if specific_files is not None:
274
self.specific_files = sorted(
275
minimum_path_selection(specific_files))
277
self.specific_files = None
278
self.specific_file_ids = None
279
self.allow_pointless = allow_pointless
280
self.revprops = revprops
281
self.message_callback = message_callback
282
self.timestamp = timestamp
283
self.timezone = timezone
284
self.committer = committer
286
self.verbose = verbose
287
# accumulates an inventory delta to the basis entry, so we can make
288
# just the necessary updates to the workingtree's cached basis.
289
self._basis_delta = []
291
self.work_tree.lock_write()
292
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
293
self.basis_revid = self.work_tree.last_revision()
294
self.basis_tree = self.work_tree.basis_tree()
295
self.basis_tree.lock_read()
297
# Cannot commit with conflicts present.
298
if len(self.work_tree.conflicts()) > 0:
299
raise ConflictsInTree
301
# Setup the bound branch variables as needed.
302
self._check_bound_branch(possible_master_transports)
304
# Check that the working tree is up to date
305
old_revno, new_revno = self._check_out_of_date_tree()
307
# Complete configuration setup
308
if reporter is not None:
309
self.reporter = reporter
310
elif self.reporter is None:
311
self.reporter = self._select_reporter()
312
if self.config is None:
313
self.config = self.branch.get_config()
315
# If provided, ensure the specified files are versioned
316
if self.specific_files is not None:
317
# Note: This routine is being called because it raises
318
# PathNotVersionedError as a side effect of finding the IDs. We
319
# later use the ids we found as input to the working tree
320
# inventory iterator, so we only consider those ids rather than
321
# examining the whole tree again.
322
# XXX: Dont we have filter_unversioned to do this more
324
self.specific_file_ids = tree.find_ids_across_trees(
325
specific_files, [self.basis_tree, self.work_tree])
327
# Setup the progress bar. As the number of files that need to be
328
# committed in unknown, progress is reported as stages.
329
# We keep track of entries separately though and include that
330
# information in the progress bar during the relevant stages.
331
self.pb_stage_name = ""
332
self.pb_stage_count = 0
333
self.pb_stage_total = 5
334
if self.bound_branch:
335
self.pb_stage_total += 1
336
self.pb.show_pct = False
337
self.pb.show_spinner = False
338
self.pb.show_eta = False
339
self.pb.show_count = True
340
self.pb.show_bar = True
342
self.basis_inv = self.basis_tree.inventory
343
self._gather_parents()
344
# After a merge, a selected file commit is not supported.
345
# See 'bzr help merge' for an explanation as to why.
346
if len(self.parents) > 1 and self.specific_files:
347
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
348
# Excludes are a form of selected file commit.
349
if len(self.parents) > 1 and self.exclude:
350
raise errors.CannotCommitSelectedFileMerge(self.exclude)
352
# Collect the changes
353
self._set_progress_stage("Collecting changes",
354
entries_title="Directory")
355
self.builder = self.branch.get_commit_builder(self.parents,
356
self.config, timestamp, timezone, committer, revprops, rev_id)
359
# find the location being committed to
360
if self.bound_branch:
361
master_location = self.master_branch.base
363
master_location = self.branch.base
365
# report the start of the commit
366
self.reporter.started(new_revno, self.rev_id, master_location)
368
self._update_builder_with_changes()
369
self._report_and_accumulate_deletes()
370
self._check_pointless()
372
# TODO: Now the new inventory is known, check for conflicts.
373
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
374
# weave lines, because nothing should be recorded until it is known
375
# that commit will succeed.
376
self._set_progress_stage("Saving data locally")
377
self.builder.finish_inventory()
379
# Prompt the user for a commit message if none provided
380
message = message_callback(self)
381
self.message = message
382
self._escape_commit_message()
384
# Add revision data to the local branch
385
self.rev_id = self.builder.commit(self.message)
391
self._process_pre_hooks(old_revno, new_revno)
393
# Upload revision data to the master.
394
# this will propagate merged revisions too if needed.
395
if self.bound_branch:
396
if not self.master_branch.repository.has_same_location(
397
self.branch.repository):
398
self._set_progress_stage("Uploading data to master branch")
399
self.master_branch.repository.fetch(self.branch.repository,
400
revision_id=self.rev_id)
401
# now the master has the revision data
402
# 'commit' to the master first so a timeout here causes the
403
# local branch to be out of date
404
self.master_branch.set_last_revision_info(new_revno,
407
# and now do the commit locally.
408
self.branch.set_last_revision_info(new_revno, self.rev_id)
410
# Make the working tree up to date with the branch
411
self._set_progress_stage("Updating the working tree")
412
self.work_tree.update_basis_by_delta(self.rev_id,
414
self.reporter.completed(new_revno, self.rev_id)
415
self._process_post_hooks(old_revno, new_revno)
420
def _select_reporter(self):
421
"""Select the CommitReporter to use."""
423
return NullCommitReporter()
424
return ReportCommitToLog()
426
def _check_pointless(self):
427
if self.allow_pointless:
429
# A merge with no effect on files
430
if len(self.parents) > 1:
432
# TODO: we could simplify this by using self._basis_delta.
434
# The initial commit adds a root directory, but this in itself is not
435
# a worthwhile commit.
436
if (self.basis_revid == revision.NULL_REVISION and
437
len(self.builder.new_inventory) == 1):
438
raise PointlessCommit()
439
# If length == 1, then we only have the root entry. Which means
440
# that there is no real difference (only the root could be different)
441
# unless deletes occured, in which case the length is irrelevant.
442
if (self.any_entries_deleted or
443
(len(self.builder.new_inventory) != 1 and
444
self.any_entries_changed)):
446
raise PointlessCommit()
448
def _check_bound_branch(self, possible_master_transports=None):
449
"""Check to see if the local branch is bound.
451
If it is bound, then most of the commit will actually be
452
done using the remote branch as the target branch.
453
Only at the end will the local branch be updated.
455
if self.local and not self.branch.get_bound_location():
456
raise errors.LocalRequiresBoundBranch()
459
self.master_branch = self.branch.get_master_branch(
460
possible_master_transports)
462
if not self.master_branch:
463
# make this branch the reference branch for out of date checks.
464
self.master_branch = self.branch
467
# If the master branch is bound, we must fail
468
master_bound_location = self.master_branch.get_bound_location()
469
if master_bound_location:
470
raise errors.CommitToDoubleBoundBranch(self.branch,
471
self.master_branch, master_bound_location)
473
# TODO: jam 20051230 We could automatically push local
474
# commits to the remote branch if they would fit.
475
# But for now, just require remote to be identical
478
# Make sure the local branch is identical to the master
479
master_info = self.master_branch.last_revision_info()
480
local_info = self.branch.last_revision_info()
481
if local_info != master_info:
482
raise errors.BoundBranchOutOfDate(self.branch,
485
# Now things are ready to change the master branch
487
self.bound_branch = self.branch
488
self.master_branch.lock_write()
489
self.master_locked = True
491
def _check_out_of_date_tree(self):
492
"""Check that the working tree is up to date.
494
:return: old_revision_number,new_revision_number tuple
497
first_tree_parent = self.work_tree.get_parent_ids()[0]
499
# if there are no parents, treat our parent as 'None'
500
# this is so that we still consider the master branch
501
# - in a checkout scenario the tree may have no
502
# parents but the branch may do.
503
first_tree_parent = bzrlib.revision.NULL_REVISION
504
old_revno, master_last = self.master_branch.last_revision_info()
505
if master_last != first_tree_parent:
506
if master_last != bzrlib.revision.NULL_REVISION:
507
raise errors.OutOfDateTree(self.work_tree)
508
if self.branch.repository.has_revision(first_tree_parent):
509
new_revno = old_revno + 1
511
# ghost parents never appear in revision history.
513
return old_revno,new_revno
515
def _process_pre_hooks(self, old_revno, new_revno):
516
"""Process any registered pre commit hooks."""
517
self._set_progress_stage("Running pre_commit hooks")
518
self._process_hooks("pre_commit", old_revno, new_revno)
520
def _process_post_hooks(self, old_revno, new_revno):
521
"""Process any registered post commit hooks."""
522
# Process the post commit hooks, if any
523
self._set_progress_stage("Running post_commit hooks")
524
# old style commit hooks - should be deprecated ? (obsoleted in
526
if self.config.post_commit() is not None:
527
hooks = self.config.post_commit().split(' ')
528
# this would be nicer with twisted.python.reflect.namedAny
530
result = eval(hook + '(branch, rev_id)',
531
{'branch':self.branch,
533
'rev_id':self.rev_id})
534
# process new style post commit hooks
535
self._process_hooks("post_commit", old_revno, new_revno)
537
def _process_hooks(self, hook_name, old_revno, new_revno):
538
if not Branch.hooks[hook_name]:
541
# new style commit hooks:
542
if not self.bound_branch:
543
hook_master = self.branch
546
hook_master = self.master_branch
547
hook_local = self.branch
548
# With bound branches, when the master is behind the local branch,
549
# the 'old_revno' and old_revid values here are incorrect.
550
# XXX: FIXME ^. RBC 20060206
552
old_revid = self.parents[0]
554
old_revid = bzrlib.revision.NULL_REVISION
556
if hook_name == "pre_commit":
557
future_tree = self.builder.revision_tree()
558
tree_delta = future_tree.changes_from(self.basis_tree,
561
for hook in Branch.hooks[hook_name]:
562
# show the running hook in the progress bar. As hooks may
563
# end up doing nothing (e.g. because they are not configured by
564
# the user) this is still showing progress, not showing overall
565
# actions - its up to each plugin to show a UI if it want's to
566
# (such as 'Emailing diff to foo@example.com').
567
self.pb_stage_name = "Running %s hooks [%s]" % \
568
(hook_name, Branch.hooks.get_hook_name(hook))
569
self._emit_progress()
570
if 'hooks' in debug.debug_flags:
571
mutter("Invoking commit hook: %r", hook)
572
if hook_name == "post_commit":
573
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
575
elif hook_name == "pre_commit":
576
hook(hook_local, hook_master,
577
old_revno, old_revid, new_revno, self.rev_id,
578
tree_delta, future_tree)
581
"""Cleanup any open locks, progress bars etc."""
582
cleanups = [self._cleanup_bound_branch,
583
self.basis_tree.unlock,
584
self.work_tree.unlock,
586
found_exception = None
587
for cleanup in cleanups:
590
# we want every cleanup to run no matter what.
591
# so we have a catchall here, but we will raise the
592
# last encountered exception up the stack: and
593
# typically this will be useful enough.
596
if found_exception is not None:
597
# don't do a plan raise, because the last exception may have been
598
# trashed, e is our sure-to-work exception even though it loses the
599
# full traceback. XXX: RBC 20060421 perhaps we could check the
600
# exc_info and if its the same one do a plain raise otherwise
601
# 'raise e' as we do now.
604
def _cleanup_bound_branch(self):
605
"""Executed at the end of a try/finally to cleanup a bound branch.
607
If the branch wasn't bound, this is a no-op.
608
If it was, it resents self.branch to the local branch, instead
611
if not self.bound_branch:
613
if self.master_locked:
614
self.master_branch.unlock()
616
def _escape_commit_message(self):
617
"""Replace xml-incompatible control characters."""
618
# FIXME: RBC 20060419 this should be done by the revision
619
# serialiser not by commit. Then we can also add an unescaper
620
# in the deserializer and start roundtripping revision messages
621
# precisely. See repository_implementations/test_repository.py
623
# Python strings can include characters that can't be
624
# represented in well-formed XML; escape characters that
625
# aren't listed in the XML specification
626
# (http://www.w3.org/TR/REC-xml/#NT-Char).
627
self.message, escape_count = re.subn(
628
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
629
lambda match: match.group(0).encode('unicode_escape'),
632
self.reporter.escaped(escape_count, self.message)
634
def _gather_parents(self):
635
"""Record the parents of a merge for merge detection."""
636
# TODO: Make sure that this list doesn't contain duplicate
637
# entries and the order is preserved when doing this.
638
self.parents = self.work_tree.get_parent_ids()
639
self.parent_invs = [self.basis_inv]
640
for revision in self.parents[1:]:
641
if self.branch.repository.has_revision(revision):
642
mutter('commit parent revision {%s}', revision)
643
inventory = self.branch.repository.get_inventory(revision)
644
self.parent_invs.append(inventory)
646
mutter('commit parent ghost revision {%s}', revision)
648
def _update_builder_with_changes(self):
649
"""Update the commit builder with the data about what has changed.
651
# Build the revision inventory.
653
# This starts by creating a new empty inventory. Depending on
654
# which files are selected for commit, and what is present in the
655
# current tree, the new inventory is populated. inventory entries
656
# which are candidates for modification have their revision set to
657
# None; inventory entries that are carried over untouched have their
658
# revision set to their prior value.
660
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
661
# results to create a new inventory at the same time, which results
662
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
665
exclude = self.exclude
666
specific_files = self.specific_files or []
667
mutter("Selecting files for commit with filter %s", specific_files)
669
# Build the new inventory
670
self._populate_from_inventory()
672
# If specific files are selected, then all un-selected files must be
673
# recorded in their previous state. For more details, see
674
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
675
if specific_files or exclude:
676
for path, old_ie in self.basis_inv.iter_entries():
677
if old_ie.file_id in self.builder.new_inventory:
678
# already added - skip.
680
if (is_inside_any(specific_files, path)
681
and not is_inside_any(exclude, path)):
682
# was inside the selected path, and not excluded - if not
683
# present it has been deleted so skip.
685
# From here down it was either not selected, or was excluded:
686
if old_ie.kind == 'directory':
687
self._next_progress_entry()
688
# We preserve the entry unaltered.
690
# Note: specific file commits after a merge are currently
691
# prohibited. This test is for sanity/safety in case it's
692
# required after that changes.
693
if len(self.parents) > 1:
695
delta, version_recorded, _ = self.builder.record_entry_contents(
696
ie, self.parent_invs, path, self.basis_tree, None)
698
self.any_entries_changed = True
700
self._basis_delta.append(delta)
702
def _report_and_accumulate_deletes(self):
703
# XXX: Could the list of deleted paths and ids be instead taken from
704
# _populate_from_inventory?
705
deleted_ids = set(self.basis_inv._byid.keys()) - \
706
set(self.builder.new_inventory._byid.keys())
708
self.any_entries_deleted = True
709
deleted = [(self.basis_tree.id2path(file_id), file_id)
710
for file_id in deleted_ids]
712
# XXX: this is not quite directory-order sorting
713
for path, file_id in deleted:
714
self._basis_delta.append((path, None, file_id, None))
715
self.reporter.deleted(path)
717
def _populate_from_inventory(self):
718
"""Populate the CommitBuilder by walking the working tree inventory."""
720
# raise an exception as soon as we find a single unknown.
721
for unknown in self.work_tree.unknowns():
722
raise StrictCommitFailed()
724
specific_files = self.specific_files
725
exclude = self.exclude
726
report_changes = self.reporter.is_verbose()
728
# A tree of paths that have been deleted. E.g. if foo/bar has been
729
# deleted, then we have {'foo':{'bar':{}}}
731
# XXX: Note that entries may have the wrong kind because the entry does
732
# not reflect the status on disk.
733
work_inv = self.work_tree.inventory
734
# NB: entries will include entries within the excluded ids/paths
735
# because iter_entries_by_dir has no 'exclude' facility today.
736
entries = work_inv.iter_entries_by_dir(
737
specific_file_ids=self.specific_file_ids, yield_parents=True)
738
for path, existing_ie in entries:
739
file_id = existing_ie.file_id
740
name = existing_ie.name
741
parent_id = existing_ie.parent_id
742
kind = existing_ie.kind
743
if kind == 'directory':
744
self._next_progress_entry()
745
# Skip files that have been deleted from the working tree.
746
# The deleted path ids are also recorded so they can be explicitly
749
path_segments = splitpath(path)
750
deleted_dict = deleted_paths
751
for segment in path_segments:
752
deleted_dict = deleted_dict.get(segment, None)
754
# We either took a path not present in the dict
755
# (deleted_dict was None), or we've reached an empty
756
# child dir in the dict, so are now a sub-path.
760
if deleted_dict is not None:
761
# the path has a deleted parent, do not add it.
763
if exclude and is_inside_any(exclude, path):
764
# Skip excluded paths. Excluded paths are processed by
765
# _update_builder_with_changes.
767
content_summary = self.work_tree.path_content_summary(path)
768
# Note that when a filter of specific files is given, we must only
769
# skip/record deleted files matching that filter.
770
if not specific_files or is_inside_any(specific_files, path):
771
if content_summary[0] == 'missing':
772
if not deleted_paths:
773
# path won't have been split yet.
774
path_segments = splitpath(path)
775
deleted_dict = deleted_paths
776
for segment in path_segments:
777
deleted_dict = deleted_dict.setdefault(segment, {})
778
self.reporter.missing(path)
779
deleted_ids.append(file_id)
781
# TODO: have the builder do the nested commit just-in-time IF and
783
if content_summary[0] == 'tree-reference':
784
# enforce repository nested tree policy.
785
if (not self.work_tree.supports_tree_reference() or
786
# repository does not support it either.
787
not self.branch.repository._format.supports_tree_reference):
788
content_summary = ('directory',) + content_summary[1:]
789
kind = content_summary[0]
790
# TODO: specific_files filtering before nested tree processing
791
if kind == 'tree-reference':
792
if self.recursive == 'down':
793
nested_revision_id = self._commit_nested_tree(
795
content_summary = content_summary[:3] + (
798
content_summary = content_summary[:3] + (
799
self.work_tree.get_reference_revision(file_id),)
801
# Record an entry for this item
802
# Note: I don't particularly want to have the existing_ie
803
# parameter but the test suite currently (28-Jun-07) breaks
804
# without it thanks to a unicode normalisation issue. :-(
805
definitely_changed = kind != existing_ie.kind
806
self._record_entry(path, file_id, specific_files, kind, name,
807
parent_id, definitely_changed, existing_ie, report_changes,
810
# Unversion IDs that were found to be deleted
811
self.work_tree.unversion(deleted_ids)
813
def _commit_nested_tree(self, file_id, path):
814
"Commit a nested tree."
815
sub_tree = self.work_tree.get_nested_tree(file_id, path)
816
# FIXME: be more comprehensive here:
817
# this works when both trees are in --trees repository,
818
# but when both are bound to a different repository,
819
# it fails; a better way of approaching this is to
820
# finally implement the explicit-caches approach design
821
# a while back - RBC 20070306.
822
if sub_tree.branch.repository.has_same_location(
823
self.work_tree.branch.repository):
824
sub_tree.branch.repository = \
825
self.work_tree.branch.repository
827
return sub_tree.commit(message=None, revprops=self.revprops,
828
recursive=self.recursive,
829
message_callback=self.message_callback,
830
timestamp=self.timestamp, timezone=self.timezone,
831
committer=self.committer,
832
allow_pointless=self.allow_pointless,
833
strict=self.strict, verbose=self.verbose,
834
local=self.local, reporter=self.reporter)
835
except errors.PointlessCommit:
836
return self.work_tree.get_reference_revision(file_id)
838
def _record_entry(self, path, file_id, specific_files, kind, name,
839
parent_id, definitely_changed, existing_ie, report_changes,
841
"Record the new inventory entry for a path if any."
842
# mutter('check %s {%s}', path, file_id)
843
# mutter('%s selected for commit', path)
844
if definitely_changed or existing_ie is None:
845
ie = make_entry(kind, name, parent_id, file_id)
847
ie = existing_ie.copy()
849
# For carried over entries we don't care about the fs hash - the repo
850
# isn't generating a sha, so we're not saving computation time.
851
delta, version_recorded, fs_hash = self.builder.record_entry_contents(
852
ie, self.parent_invs, path, self.work_tree, content_summary)
854
self._basis_delta.append(delta)
856
self.any_entries_changed = True
858
self._report_change(ie, path)
860
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
863
def _report_change(self, ie, path):
864
"""Report a change to the user.
866
The change that has occurred is described relative to the basis
869
if (self.basis_inv.has_id(ie.file_id)):
870
basis_ie = self.basis_inv[ie.file_id]
873
change = ie.describe_change(basis_ie, ie)
874
if change in (InventoryEntry.RENAMED,
875
InventoryEntry.MODIFIED_AND_RENAMED):
876
old_path = self.basis_inv.id2path(ie.file_id)
877
self.reporter.renamed(change, old_path, path)
879
self.reporter.snapshot_change(change, path)
881
def _set_progress_stage(self, name, entries_title=None):
882
"""Set the progress stage and emit an update to the progress bar."""
883
self.pb_stage_name = name
884
self.pb_stage_count += 1
885
self.pb_entries_title = entries_title
886
if entries_title is not None:
887
self.pb_entries_count = 0
888
self.pb_entries_total = '?'
889
self._emit_progress()
891
def _next_progress_entry(self):
892
"""Emit an update to the progress bar and increment the entry count."""
893
self.pb_entries_count += 1
894
self._emit_progress()
896
def _emit_progress(self):
897
if self.pb_entries_title:
898
if self.pb_entries_total == '?':
899
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
900
self.pb_entries_title, self.pb_entries_count)
902
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
903
self.pb_entries_title, self.pb_entries_count,
904
str(self.pb_entries_total))
906
text = "%s - Stage" % (self.pb_stage_name)
907
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)