1
# Copyright (C) 2005 Canonical Ltd
1
# Copyright (C) 2005, 2006, 2007, 2008 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
# FIXME: "bzr commit doc/format" commits doc/format.txt!
21
def commit(branch, message,
28
allow_pointless=True):
29
"""Commit working copy as a new revision.
31
The basic approach is to add all the file texts into the
32
store, then the inventory, then make a new revision pointing
33
to that inventory and store that.
35
This is not quite safe if the working copy changes during the
36
commit; for the moment that is simply not allowed. A better
37
approach is to make a temporary copy of the files before
38
computing their hashes, and then add those hashes in turn to
39
the inventory. This should mean at least that there are no
40
broken hash pointers. There is no way we can get a snapshot
41
of the whole directory at an instant. This would also have to
42
be robust against files disappearing, moving, etc. So the
43
whole thing is a bit hard.
45
This raises PointlessCommit if there are no changes, no new merges,
46
and allow_pointless is false.
48
timestamp -- if not None, seconds-since-epoch for a
49
postdated/predated commit.
52
If true, commit only those files.
55
If set, use this as the new revision id.
56
Useful for test or import commands that need to tightly
57
control what revisions are assigned. If you duplicate
58
a revision id that exists elsewhere it is your own fault.
59
If null (default), a time/random revision id is generated.
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
66
from bzrlib.branch import Branch
68
from bzrlib.errors import (BzrError, PointlessCommit,
72
from bzrlib.osutils import (get_user_encoding,
73
kind_marker, isdir,isfile, is_inside_any,
74
is_inside_or_parent_of_any,
75
minimum_path_selection,
76
quotefn, sha_file, split_lines,
79
from bzrlib.testament import Testament
80
from bzrlib.trace import mutter, note, warning, is_quiet
81
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
82
from bzrlib import symbol_versioning
83
from bzrlib.symbol_versioning import (deprecated_passed,
86
from bzrlib.workingtree import WorkingTree
87
from bzrlib.urlutils import unescape_for_display
91
class NullCommitReporter(object):
92
"""I report on progress of a commit."""
94
def started(self, revno, revid, location=None):
96
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
97
"to started.", DeprecationWarning,
101
def snapshot_change(self, change, path):
104
def completed(self, revno, rev_id):
107
def deleted(self, file_id):
110
def escaped(self, escape_count, message):
113
def missing(self, path):
116
def renamed(self, change, old_path, new_path):
119
def is_verbose(self):
123
class ReportCommitToLog(NullCommitReporter):
125
def _note(self, format, *args):
128
Subclasses may choose to override this method.
132
def snapshot_change(self, change, path):
133
if change == 'unchanged':
135
if change == 'added' and path == '':
137
self._note("%s %s", change, path)
139
def started(self, revno, rev_id, location=None):
140
if location is not None:
141
location = ' to: ' + unescape_for_display(location, 'utf-8')
143
# When started was added, location was only made optional by
144
# accident. Matt Nordhoff 20071129
145
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
146
"to started.", DeprecationWarning,
149
self._note('Committing%s', location)
151
def completed(self, revno, rev_id):
152
self._note('Committed revision %d.', revno)
154
def deleted(self, file_id):
155
self._note('deleted %s', file_id)
157
def escaped(self, escape_count, message):
158
self._note("replaced %d control characters in message", escape_count)
160
def missing(self, path):
161
self._note('missing %s', path)
163
def renamed(self, change, old_path, new_path):
164
self._note('%s %s => %s', change, old_path, new_path)
166
def is_verbose(self):
170
class Commit(object):
171
"""Task of committing a new revision.
173
This is a MethodObject: it accumulates state as the commit is
174
prepared, and then it is discarded. It doesn't represent
175
historical revisions, just the act of recording a new one.
178
Modified to hold a list of files that have been deleted from
179
the working directory; these should be removed from the
64
from bzrlib.osutils import local_time_offset, username
65
from bzrlib.branch import gen_file_id
66
from bzrlib.errors import BzrError, PointlessCommit
67
from bzrlib.revision import Revision, RevisionReference
68
from bzrlib.trace import mutter, note
69
from bzrlib.xml import pack_xml
74
# First walk over the working inventory; and both update that
75
# and also build a new revision inventory. The revision
76
# inventory needs to hold the text-id, sha1 and size of the
77
# actual file versions committed in the revision. (These are
78
# not present in the working inventory.) We also need to
79
# detect missing/deleted files, and remove them from the
82
work_tree = branch.working_tree()
83
work_inv = work_tree.inventory
84
basis = branch.basis_tree()
85
basis_inv = basis.inventory
88
note('looking for changes...')
90
pending_merges = branch.pending_merges()
92
missing_ids, new_inv, any_changes = \
93
_gather_commit(branch,
100
if not (any_changes or allow_pointless or pending_merges):
185
"""Create a Commit object.
187
:param reporter: the default reporter to use or None to decide later
189
self.reporter = reporter
199
allow_pointless=True,
207
message_callback=None,
210
possible_master_transports=None):
211
"""Commit working copy as a new revision.
213
:param message: the commit message (it or message_callback is required)
215
:param timestamp: if not None, seconds-since-epoch for a
216
postdated/predated commit.
218
:param specific_files: If true, commit only those files.
220
:param rev_id: If set, use this as the new revision id.
221
Useful for test or import commands that need to tightly
222
control what revisions are assigned. If you duplicate
223
a revision id that exists elsewhere it is your own fault.
224
If null (default), a time/random revision id is generated.
226
:param allow_pointless: If true (default), commit even if nothing
227
has changed and no merges are recorded.
229
:param strict: If true, don't allow a commit if the working tree
230
contains unknown files.
232
:param revprops: Properties for new revision
233
:param local: Perform a local only commit.
234
:param reporter: the reporter to use or None for the default
235
:param verbose: if True and the reporter is not None, report everything
236
:param recursive: If set to 'down', commit in any subtrees that have
237
pending changes of any sort during this commit.
238
:param exclude: None or a list of relative paths to exclude from the
239
commit. Pending changes to excluded files will be ignored by the
242
mutter('preparing to commit')
244
if working_tree is None:
245
raise BzrError("working_tree must be passed into commit().")
247
self.work_tree = working_tree
248
self.branch = self.work_tree.branch
249
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
250
if not self.branch.repository.supports_rich_root():
251
raise errors.RootNotRich()
252
if message_callback is None:
253
if message is not None:
254
if isinstance(message, str):
255
message = message.decode(get_user_encoding())
256
message_callback = lambda x: message
258
raise BzrError("The message or message_callback keyword"
259
" parameter is required for commit().")
261
self.bound_branch = None
262
self.any_entries_changed = False
263
self.any_entries_deleted = False
264
if exclude is not None:
265
self.exclude = sorted(
266
minimum_path_selection(exclude))
270
self.master_branch = None
271
self.master_locked = False
272
self.recursive = recursive
274
if specific_files is not None:
275
self.specific_files = sorted(
276
minimum_path_selection(specific_files))
278
self.specific_files = None
279
self.specific_file_ids = None
280
self.allow_pointless = allow_pointless
281
self.revprops = revprops
282
self.message_callback = message_callback
283
self.timestamp = timestamp
284
self.timezone = timezone
285
self.committer = committer
287
self.verbose = verbose
289
self.work_tree.lock_write()
290
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
291
self.basis_revid = self.work_tree.last_revision()
292
self.basis_tree = self.work_tree.basis_tree()
293
self.basis_tree.lock_read()
295
# Cannot commit with conflicts present.
296
if len(self.work_tree.conflicts()) > 0:
297
raise ConflictsInTree
299
# Setup the bound branch variables as needed.
300
self._check_bound_branch(possible_master_transports)
302
# Check that the working tree is up to date
303
old_revno, new_revno = self._check_out_of_date_tree()
305
# Complete configuration setup
306
if reporter is not None:
307
self.reporter = reporter
308
elif self.reporter is None:
309
self.reporter = self._select_reporter()
310
if self.config is None:
311
self.config = self.branch.get_config()
313
# If provided, ensure the specified files are versioned
314
if self.specific_files is not None:
315
# Note: This routine is being called because it raises
316
# PathNotVersionedError as a side effect of finding the IDs. We
317
# later use the ids we found as input to the working tree
318
# inventory iterator, so we only consider those ids rather than
319
# examining the whole tree again.
320
# XXX: Dont we have filter_unversioned to do this more
322
self.specific_file_ids = tree.find_ids_across_trees(
323
specific_files, [self.basis_tree, self.work_tree])
325
# Setup the progress bar. As the number of files that need to be
326
# committed in unknown, progress is reported as stages.
327
# We keep track of entries separately though and include that
328
# information in the progress bar during the relevant stages.
329
self.pb_stage_name = ""
330
self.pb_stage_count = 0
331
self.pb_stage_total = 5
332
if self.bound_branch:
333
self.pb_stage_total += 1
334
self.pb.show_pct = False
335
self.pb.show_spinner = False
336
self.pb.show_eta = False
337
self.pb.show_count = True
338
self.pb.show_bar = True
340
self.basis_inv = self.basis_tree.inventory
341
self._gather_parents()
342
# After a merge, a selected file commit is not supported.
343
# See 'bzr help merge' for an explanation as to why.
344
if len(self.parents) > 1 and self.specific_files:
345
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
346
# Excludes are a form of selected file commit.
347
if len(self.parents) > 1 and self.exclude:
348
raise errors.CannotCommitSelectedFileMerge(self.exclude)
350
# Collect the changes
351
self._set_progress_stage("Collecting changes",
352
entries_title="Directory")
353
self.builder = self.branch.get_commit_builder(self.parents,
354
self.config, timestamp, timezone, committer, revprops, rev_id)
357
self.builder.will_record_deletes()
358
# find the location being committed to
359
if self.bound_branch:
360
master_location = self.master_branch.base
362
master_location = self.branch.base
364
# report the start of the commit
365
self.reporter.started(new_revno, self.rev_id, master_location)
367
self._update_builder_with_changes()
368
self._report_and_accumulate_deletes()
369
self._check_pointless()
371
# TODO: Now the new inventory is known, check for conflicts.
372
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
373
# weave lines, because nothing should be recorded until it is known
374
# that commit will succeed.
375
self._set_progress_stage("Saving data locally")
376
self.builder.finish_inventory()
378
# Prompt the user for a commit message if none provided
379
message = message_callback(self)
380
self.message = message
381
self._escape_commit_message()
383
# Add revision data to the local branch
384
self.rev_id = self.builder.commit(self.message)
387
mutter("aborting commit write group because of exception:")
388
trace.log_exception_quietly()
389
note("aborting commit write group: %r" % (e,))
393
self._process_pre_hooks(old_revno, new_revno)
395
# Upload revision data to the master.
396
# this will propagate merged revisions too if needed.
397
if self.bound_branch:
398
if not self.master_branch.repository.has_same_location(
399
self.branch.repository):
400
self._set_progress_stage("Uploading data to master branch")
401
self.master_branch.repository.fetch(self.branch.repository,
402
revision_id=self.rev_id)
403
# now the master has the revision data
404
# 'commit' to the master first so a timeout here causes the
405
# local branch to be out of date
406
self.master_branch.set_last_revision_info(new_revno,
409
# and now do the commit locally.
410
self.branch.set_last_revision_info(new_revno, self.rev_id)
412
# Make the working tree up to date with the branch
413
self._set_progress_stage("Updating the working tree")
414
self.work_tree.update_basis_by_delta(self.rev_id,
415
self.builder.get_basis_delta())
416
self.reporter.completed(new_revno, self.rev_id)
417
self._process_post_hooks(old_revno, new_revno)
422
def _select_reporter(self):
423
"""Select the CommitReporter to use."""
425
return NullCommitReporter()
426
return ReportCommitToLog()
428
def _check_pointless(self):
429
if self.allow_pointless:
431
# A merge with no effect on files
432
if len(self.parents) > 1:
434
# TODO: we could simplify this by using self.builder.basis_delta.
436
# The initial commit adds a root directory, but this in itself is not
437
# a worthwhile commit.
438
if (self.basis_revid == revision.NULL_REVISION and
439
len(self.builder.new_inventory) == 1):
101
440
raise PointlessCommit()
103
for file_id in missing_ids:
104
# Any files that have been deleted are now removed from the
105
# working inventory. Files that were not selected for commit
106
# are left as they were in the working inventory and ommitted
107
# from the revision inventory.
109
# have to do this later so we don't mess up the iterator.
110
# since parents may be removed before their children we
113
# FIXME: There's probably a better way to do this; perhaps
114
# the workingtree should know how to filter itbranch.
115
if work_inv.has_id(file_id):
116
del work_inv[file_id]
120
rev_id = _gen_revision_id(time.time())
123
inv_tmp = tempfile.TemporaryFile()
124
pack_xml(new_inv, inv_tmp)
126
branch.inventory_store.add(inv_tmp, inv_id)
127
mutter('new inventory_id is {%s}' % inv_id)
129
# We could also just sha hash the inv_tmp file
130
# however, in the case that branch.inventory_store.add()
131
# ever actually does anything special
132
inv_sha1 = branch.get_inventory_sha1(inv_id)
134
branch._write_inventory(work_inv)
136
if timestamp == None:
137
timestamp = time.time()
139
if committer == None:
140
committer = username()
143
timezone = local_time_offset()
145
mutter("building commit log message")
146
rev = Revision(timestamp=timestamp,
151
inventory_sha1=inv_sha1,
155
precursor_id = branch.last_patch()
157
precursor_sha1 = branch.get_revision_sha1(precursor_id)
158
rev.parents.append(RevisionReference(precursor_id, precursor_sha1))
159
for merge_rev in pending_merges:
160
rev.parents.append(RevisionReference(merge_rev))
162
rev_tmp = tempfile.TemporaryFile()
163
pack_xml(rev, rev_tmp)
165
branch.revision_store.add(rev_tmp, rev_id)
166
mutter("new revision_id is {%s}" % rev_id)
168
## XXX: Everything up to here can simply be orphaned if we abort
169
## the commit; it will leave junk files behind but that doesn't
172
## TODO: Read back the just-generated changeset, and make sure it
173
## applies and recreates the right state.
175
## TODO: Also calculate and store the inventory SHA1
176
mutter("committing patch r%d" % (branch.revno() + 1))
178
branch.append_revision(rev_id)
180
branch.set_pending_merges([])
183
note("commited r%d" % branch.revno())
189
def _gen_revision_id(when):
190
"""Return new revision-id."""
191
from binascii import hexlify
192
from osutils import rand_bytes, compact_date, user_email
194
s = '%s-%s-' % (user_email(), compact_date(when))
195
s += hexlify(rand_bytes(8))
199
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
201
"""Build inventory preparatory to commit.
203
Returns missing_ids, new_inv, any_changes.
205
This adds any changed files into the text store, and sets their
206
test-id, sha and size in the returned inventory appropriately.
209
Modified to hold a list of files that have been deleted from
210
the working directory; these should be removed from the
213
from bzrlib.inventory import Inventory
214
from bzrlib.osutils import isdir, isfile, sha_string, quotefn, \
215
local_time_offset, username, kind_marker, is_inside_any
217
from bzrlib.branch import gen_file_id
218
from bzrlib.errors import BzrError
219
from bzrlib.revision import Revision
220
from bzrlib.trace import mutter, note
223
inv = Inventory(work_inv.root.file_id)
226
for path, entry in work_inv.iter_entries():
227
## TODO: Check that the file kind has not changed from the previous
228
## revision of this file (if any).
230
p = branch.abspath(path)
231
file_id = entry.file_id
232
mutter('commit prep file %s, id %r ' % (p, file_id))
234
if specific_files and not is_inside_any(specific_files, path):
235
mutter(' skipping file excluded from commit')
236
if basis_inv.has_id(file_id):
237
# carry over with previous state
238
inv.add(basis_inv[file_id].copy())
240
# omit this from committed inventory
244
if not work_tree.has_id(file_id):
246
print('deleted %s%s' % (path, kind_marker(entry.kind)))
248
mutter(" file is missing, removing from inventory")
249
missing_ids.append(file_id)
252
# this is present in the new inventory; may be new, modified or
254
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
260
old_kind = old_ie.kind
261
if old_kind != entry.kind:
262
raise BzrError("entry %r changed kind from %r to %r"
263
% (file_id, old_kind, entry.kind))
265
if entry.kind == 'directory':
267
raise BzrError("%s is entered as directory but not a directory"
269
elif entry.kind == 'file':
271
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
273
new_sha1 = work_tree.get_file_sha1(file_id)
276
and old_ie.text_sha1 == new_sha1):
277
## assert content == basis.get_file(file_id).read()
278
entry.text_id = old_ie.text_id
279
entry.text_sha1 = new_sha1
280
entry.text_size = old_ie.text_size
281
mutter(' unchanged from previous text_id {%s}' %
284
content = file(p, 'rb').read()
286
# calculate the sha again, just in case the file contents
287
# changed since we updated the cache
288
entry.text_sha1 = sha_string(content)
289
entry.text_size = len(content)
291
entry.text_id = gen_file_id(entry.name)
292
branch.text_store.add(content, entry.text_id)
293
mutter(' stored with text_id {%s}' % entry.text_id)
296
marked = path + kind_marker(entry.kind)
298
print 'added', marked
300
elif old_ie == entry:
302
elif (old_ie.name == entry.name
303
and old_ie.parent_id == entry.parent_id):
304
print 'modified', marked
307
print 'renamed', marked
310
return missing_ids, inv, any_changes
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
# unless deletes occured, in which case the length is irrelevant.
444
if (self.any_entries_deleted or
445
(len(self.builder.new_inventory) != 1 and
446
self.any_entries_changed)):
448
raise PointlessCommit()
450
def _check_bound_branch(self, possible_master_transports=None):
451
"""Check to see if the local branch is bound.
453
If it is bound, then most of the commit will actually be
454
done using the remote branch as the target branch.
455
Only at the end will the local branch be updated.
457
if self.local and not self.branch.get_bound_location():
458
raise errors.LocalRequiresBoundBranch()
461
self.master_branch = self.branch.get_master_branch(
462
possible_master_transports)
464
if not self.master_branch:
465
# make this branch the reference branch for out of date checks.
466
self.master_branch = self.branch
469
# If the master branch is bound, we must fail
470
master_bound_location = self.master_branch.get_bound_location()
471
if master_bound_location:
472
raise errors.CommitToDoubleBoundBranch(self.branch,
473
self.master_branch, master_bound_location)
475
# TODO: jam 20051230 We could automatically push local
476
# commits to the remote branch if they would fit.
477
# But for now, just require remote to be identical
480
# Make sure the local branch is identical to the master
481
master_info = self.master_branch.last_revision_info()
482
local_info = self.branch.last_revision_info()
483
if local_info != master_info:
484
raise errors.BoundBranchOutOfDate(self.branch,
487
# Now things are ready to change the master branch
489
self.bound_branch = self.branch
490
self.master_branch.lock_write()
491
self.master_locked = True
493
def _check_out_of_date_tree(self):
494
"""Check that the working tree is up to date.
496
:return: old_revision_number,new_revision_number tuple
499
first_tree_parent = self.work_tree.get_parent_ids()[0]
501
# if there are no parents, treat our parent as 'None'
502
# this is so that we still consider the master branch
503
# - in a checkout scenario the tree may have no
504
# parents but the branch may do.
505
first_tree_parent = bzrlib.revision.NULL_REVISION
506
old_revno, master_last = self.master_branch.last_revision_info()
507
if master_last != first_tree_parent:
508
if master_last != bzrlib.revision.NULL_REVISION:
509
raise errors.OutOfDateTree(self.work_tree)
510
if self.branch.repository.has_revision(first_tree_parent):
511
new_revno = old_revno + 1
513
# ghost parents never appear in revision history.
515
return old_revno,new_revno
517
def _process_pre_hooks(self, old_revno, new_revno):
518
"""Process any registered pre commit hooks."""
519
self._set_progress_stage("Running pre_commit hooks")
520
self._process_hooks("pre_commit", old_revno, new_revno)
522
def _process_post_hooks(self, old_revno, new_revno):
523
"""Process any registered post commit hooks."""
524
# Process the post commit hooks, if any
525
self._set_progress_stage("Running post_commit hooks")
526
# old style commit hooks - should be deprecated ? (obsoleted in
528
if self.config.post_commit() is not None:
529
hooks = self.config.post_commit().split(' ')
530
# this would be nicer with twisted.python.reflect.namedAny
532
result = eval(hook + '(branch, rev_id)',
533
{'branch':self.branch,
535
'rev_id':self.rev_id})
536
# process new style post commit hooks
537
self._process_hooks("post_commit", old_revno, new_revno)
539
def _process_hooks(self, hook_name, old_revno, new_revno):
540
if not Branch.hooks[hook_name]:
543
# new style commit hooks:
544
if not self.bound_branch:
545
hook_master = self.branch
548
hook_master = self.master_branch
549
hook_local = self.branch
550
# With bound branches, when the master is behind the local branch,
551
# the 'old_revno' and old_revid values here are incorrect.
552
# XXX: FIXME ^. RBC 20060206
554
old_revid = self.parents[0]
556
old_revid = bzrlib.revision.NULL_REVISION
558
if hook_name == "pre_commit":
559
future_tree = self.builder.revision_tree()
560
tree_delta = future_tree.changes_from(self.basis_tree,
563
for hook in Branch.hooks[hook_name]:
564
# show the running hook in the progress bar. As hooks may
565
# end up doing nothing (e.g. because they are not configured by
566
# the user) this is still showing progress, not showing overall
567
# actions - its up to each plugin to show a UI if it want's to
568
# (such as 'Emailing diff to foo@example.com').
569
self.pb_stage_name = "Running %s hooks [%s]" % \
570
(hook_name, Branch.hooks.get_hook_name(hook))
571
self._emit_progress()
572
if 'hooks' in debug.debug_flags:
573
mutter("Invoking commit hook: %r", hook)
574
if hook_name == "post_commit":
575
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
577
elif hook_name == "pre_commit":
578
hook(hook_local, hook_master,
579
old_revno, old_revid, new_revno, self.rev_id,
580
tree_delta, future_tree)
583
"""Cleanup any open locks, progress bars etc."""
584
cleanups = [self._cleanup_bound_branch,
585
self.basis_tree.unlock,
586
self.work_tree.unlock,
588
found_exception = None
589
for cleanup in cleanups:
592
# we want every cleanup to run no matter what.
593
# so we have a catchall here, but we will raise the
594
# last encountered exception up the stack: and
595
# typically this will be useful enough.
598
if found_exception is not None:
599
# don't do a plan raise, because the last exception may have been
600
# trashed, e is our sure-to-work exception even though it loses the
601
# full traceback. XXX: RBC 20060421 perhaps we could check the
602
# exc_info and if its the same one do a plain raise otherwise
603
# 'raise e' as we do now.
606
def _cleanup_bound_branch(self):
607
"""Executed at the end of a try/finally to cleanup a bound branch.
609
If the branch wasn't bound, this is a no-op.
610
If it was, it resents self.branch to the local branch, instead
613
if not self.bound_branch:
615
if self.master_locked:
616
self.master_branch.unlock()
618
def _escape_commit_message(self):
619
"""Replace xml-incompatible control characters."""
620
# FIXME: RBC 20060419 this should be done by the revision
621
# serialiser not by commit. Then we can also add an unescaper
622
# in the deserializer and start roundtripping revision messages
623
# precisely. See repository_implementations/test_repository.py
625
# Python strings can include characters that can't be
626
# represented in well-formed XML; escape characters that
627
# aren't listed in the XML specification
628
# (http://www.w3.org/TR/REC-xml/#NT-Char).
629
self.message, escape_count = re.subn(
630
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
631
lambda match: match.group(0).encode('unicode_escape'),
634
self.reporter.escaped(escape_count, self.message)
636
def _gather_parents(self):
637
"""Record the parents of a merge for merge detection."""
638
# TODO: Make sure that this list doesn't contain duplicate
639
# entries and the order is preserved when doing this.
640
self.parents = self.work_tree.get_parent_ids()
641
self.parent_invs = [self.basis_inv]
642
for revision in self.parents[1:]:
643
if self.branch.repository.has_revision(revision):
644
mutter('commit parent revision {%s}', revision)
645
inventory = self.branch.repository.get_inventory(revision)
646
self.parent_invs.append(inventory)
648
mutter('commit parent ghost revision {%s}', revision)
650
def _update_builder_with_changes(self):
651
"""Update the commit builder with the data about what has changed.
653
# Build the revision inventory.
655
# This starts by creating a new empty inventory. Depending on
656
# which files are selected for commit, and what is present in the
657
# current tree, the new inventory is populated. inventory entries
658
# which are candidates for modification have their revision set to
659
# None; inventory entries that are carried over untouched have their
660
# revision set to their prior value.
662
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
663
# results to create a new inventory at the same time, which results
664
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
667
exclude = self.exclude
668
specific_files = self.specific_files or []
669
mutter("Selecting files for commit with filter %s", specific_files)
671
# Build the new inventory
672
self._populate_from_inventory()
674
# If specific files are selected, then all un-selected files must be
675
# recorded in their previous state. For more details, see
676
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
677
if specific_files or exclude:
678
for path, old_ie in self.basis_inv.iter_entries():
679
if old_ie.file_id in self.builder.new_inventory:
680
# already added - skip.
682
if (is_inside_any(specific_files, path)
683
and not is_inside_any(exclude, path)):
684
# was inside the selected path, and not excluded - if not
685
# present it has been deleted so skip.
687
# From here down it was either not selected, or was excluded:
688
if old_ie.kind == 'directory':
689
self._next_progress_entry()
690
# We preserve the entry unaltered.
692
# Note: specific file commits after a merge are currently
693
# prohibited. This test is for sanity/safety in case it's
694
# required after that changes.
695
if len(self.parents) > 1:
697
_, version_recorded, _ = self.builder.record_entry_contents(
698
ie, self.parent_invs, path, self.basis_tree, None)
700
self.any_entries_changed = True
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
if (isinstance(self.basis_inv, Inventory)
706
and isinstance(self.builder.new_inventory, Inventory)):
707
# the older Inventory classes provide a _byid dict, and building a
708
# set from the keys of this dict is substantially faster than even
709
# getting a set of ids from the inventory
711
# <lifeless> set(dict) is roughly the same speed as
712
# set(iter(dict)) and both are significantly slower than
714
deleted_ids = set(self.basis_inv._byid.keys()) - \
715
set(self.builder.new_inventory._byid.keys())
717
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
719
self.any_entries_deleted = True
720
deleted = [(self.basis_tree.id2path(file_id), file_id)
721
for file_id in deleted_ids]
723
# XXX: this is not quite directory-order sorting
724
for path, file_id in deleted:
725
self.builder.record_delete(path, file_id)
726
self.reporter.deleted(path)
728
def _populate_from_inventory(self):
729
"""Populate the CommitBuilder by walking the working tree inventory."""
731
# raise an exception as soon as we find a single unknown.
732
for unknown in self.work_tree.unknowns():
733
raise StrictCommitFailed()
735
specific_files = self.specific_files
736
exclude = self.exclude
737
report_changes = self.reporter.is_verbose()
739
# A tree of paths that have been deleted. E.g. if foo/bar has been
740
# deleted, then we have {'foo':{'bar':{}}}
742
# XXX: Note that entries may have the wrong kind because the entry does
743
# not reflect the status on disk.
744
work_inv = self.work_tree.inventory
745
# NB: entries will include entries within the excluded ids/paths
746
# because iter_entries_by_dir has no 'exclude' facility today.
747
entries = work_inv.iter_entries_by_dir(
748
specific_file_ids=self.specific_file_ids, yield_parents=True)
749
for path, existing_ie in entries:
750
file_id = existing_ie.file_id
751
name = existing_ie.name
752
parent_id = existing_ie.parent_id
753
kind = existing_ie.kind
754
if kind == 'directory':
755
self._next_progress_entry()
756
# Skip files that have been deleted from the working tree.
757
# The deleted path ids are also recorded so they can be explicitly
760
path_segments = splitpath(path)
761
deleted_dict = deleted_paths
762
for segment in path_segments:
763
deleted_dict = deleted_dict.get(segment, None)
765
# We either took a path not present in the dict
766
# (deleted_dict was None), or we've reached an empty
767
# child dir in the dict, so are now a sub-path.
771
if deleted_dict is not None:
772
# the path has a deleted parent, do not add it.
774
if exclude and is_inside_any(exclude, path):
775
# Skip excluded paths. Excluded paths are processed by
776
# _update_builder_with_changes.
778
content_summary = self.work_tree.path_content_summary(path)
779
# Note that when a filter of specific files is given, we must only
780
# skip/record deleted files matching that filter.
781
if not specific_files or is_inside_any(specific_files, path):
782
if content_summary[0] == 'missing':
783
if not deleted_paths:
784
# path won't have been split yet.
785
path_segments = splitpath(path)
786
deleted_dict = deleted_paths
787
for segment in path_segments:
788
deleted_dict = deleted_dict.setdefault(segment, {})
789
self.reporter.missing(path)
790
deleted_ids.append(file_id)
792
# TODO: have the builder do the nested commit just-in-time IF and
794
if content_summary[0] == 'tree-reference':
795
# enforce repository nested tree policy.
796
if (not self.work_tree.supports_tree_reference() or
797
# repository does not support it either.
798
not self.branch.repository._format.supports_tree_reference):
799
content_summary = ('directory',) + content_summary[1:]
800
kind = content_summary[0]
801
# TODO: specific_files filtering before nested tree processing
802
if kind == 'tree-reference':
803
if self.recursive == 'down':
804
nested_revision_id = self._commit_nested_tree(
806
content_summary = content_summary[:3] + (
809
content_summary = content_summary[:3] + (
810
self.work_tree.get_reference_revision(file_id),)
812
# Record an entry for this item
813
# Note: I don't particularly want to have the existing_ie
814
# parameter but the test suite currently (28-Jun-07) breaks
815
# without it thanks to a unicode normalisation issue. :-(
816
definitely_changed = kind != existing_ie.kind
817
self._record_entry(path, file_id, specific_files, kind, name,
818
parent_id, definitely_changed, existing_ie, report_changes,
821
# Unversion IDs that were found to be deleted
822
self.work_tree.unversion(deleted_ids)
824
def _commit_nested_tree(self, file_id, path):
825
"Commit a nested tree."
826
sub_tree = self.work_tree.get_nested_tree(file_id, path)
827
# FIXME: be more comprehensive here:
828
# this works when both trees are in --trees repository,
829
# but when both are bound to a different repository,
830
# it fails; a better way of approaching this is to
831
# finally implement the explicit-caches approach design
832
# a while back - RBC 20070306.
833
if sub_tree.branch.repository.has_same_location(
834
self.work_tree.branch.repository):
835
sub_tree.branch.repository = \
836
self.work_tree.branch.repository
838
return sub_tree.commit(message=None, revprops=self.revprops,
839
recursive=self.recursive,
840
message_callback=self.message_callback,
841
timestamp=self.timestamp, timezone=self.timezone,
842
committer=self.committer,
843
allow_pointless=self.allow_pointless,
844
strict=self.strict, verbose=self.verbose,
845
local=self.local, reporter=self.reporter)
846
except errors.PointlessCommit:
847
return self.work_tree.get_reference_revision(file_id)
849
def _record_entry(self, path, file_id, specific_files, kind, name,
850
parent_id, definitely_changed, existing_ie, report_changes,
852
"Record the new inventory entry for a path if any."
853
# mutter('check %s {%s}', path, file_id)
854
# mutter('%s selected for commit', path)
855
if definitely_changed or existing_ie is None:
856
ie = make_entry(kind, name, parent_id, file_id)
858
ie = existing_ie.copy()
860
# For carried over entries we don't care about the fs hash - the repo
861
# isn't generating a sha, so we're not saving computation time.
862
_, version_recorded, fs_hash = self.builder.record_entry_contents(
863
ie, self.parent_invs, path, self.work_tree, content_summary)
865
self.any_entries_changed = True
867
self._report_change(ie, path)
869
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
872
def _report_change(self, ie, path):
873
"""Report a change to the user.
875
The change that has occurred is described relative to the basis
878
if (self.basis_inv.has_id(ie.file_id)):
879
basis_ie = self.basis_inv[ie.file_id]
882
change = ie.describe_change(basis_ie, ie)
883
if change in (InventoryEntry.RENAMED,
884
InventoryEntry.MODIFIED_AND_RENAMED):
885
old_path = self.basis_inv.id2path(ie.file_id)
886
self.reporter.renamed(change, old_path, path)
888
self.reporter.snapshot_change(change, path)
890
def _set_progress_stage(self, name, entries_title=None):
891
"""Set the progress stage and emit an update to the progress bar."""
892
self.pb_stage_name = name
893
self.pb_stage_count += 1
894
self.pb_entries_title = entries_title
895
if entries_title is not None:
896
self.pb_entries_count = 0
897
self.pb_entries_total = '?'
898
self._emit_progress()
900
def _next_progress_entry(self):
901
"""Emit an update to the progress bar and increment the entry count."""
902
self.pb_entries_count += 1
903
self._emit_progress()
905
def _emit_progress(self):
906
if self.pb_entries_title:
907
if self.pb_entries_total == '?':
908
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
909
self.pb_entries_title, self.pb_entries_count)
911
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
912
self.pb_entries_title, self.pb_entries_count,
913
str(self.pb_entries_total))
915
text = "%s - Stage" % (self.pb_stage_name)
916
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
added
removed
bzrlib/commit.py
