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,
26
"""Commit working copy as a new revision.
28
The basic approach is to add all the file texts into the
29
store, then the inventory, then make a new revision pointing
30
to that inventory and store that.
32
This is not quite safe if the working copy changes during the
33
commit; for the moment that is simply not allowed. A better
34
approach is to make a temporary copy of the files before
35
computing their hashes, and then add those hashes in turn to
36
the inventory. This should mean at least that there are no
37
broken hash pointers. There is no way we can get a snapshot
38
of the whole directory at an instant. This would also have to
39
be robust against files disappearing, moving, etc. So the
40
whole thing is a bit hard.
42
timestamp -- if not None, seconds-since-epoch for a
43
postdated/predated commit.
46
If true, commit only those files.
49
If set, use this as the new revision id.
50
Useful for test or import commands that need to tightly
51
control what revisions are assigned. If you duplicate
52
a revision id that exists elsewhere it is your own fault.
53
If null (default), a time/random revision id is generated.
58
from osutils import local_time_offset, username
60
from branch import gen_file_id
61
from errors import BzrError
62
from revision import Revision
63
from trace import mutter, note
68
# First walk over the working inventory; and both update that
69
# and also build a new revision inventory. The revision
70
# inventory needs to hold the text-id, sha1 and size of the
71
# actual file versions committed in the revision. (These are
72
# not present in the working inventory.) We also need to
73
# detect missing/deleted files, and remove them from the
76
work_tree = branch.working_tree()
77
work_inv = work_tree.inventory
78
basis = branch.basis_tree()
79
basis_inv = basis.inventory
82
note('looking for changes...')
84
missing_ids, new_inv = _gather_commit(branch,
91
for file_id in missing_ids:
92
# Any files that have been deleted are now removed from the
93
# working inventory. Files that were not selected for commit
94
# are left as they were in the working inventory and ommitted
95
# from the revision inventory.
97
# have to do this later so we don't mess up the iterator.
98
# since parents may be removed before their children we
101
# FIXME: There's probably a better way to do this; perhaps
102
# the workingtree should know how to filter itbranch.
103
if work_inv.has_id(file_id):
104
del work_inv[file_id]
108
rev_id = _gen_revision_id(time.time())
111
inv_tmp = tempfile.TemporaryFile()
112
new_inv.write_xml(inv_tmp)
114
branch.inventory_store.add(inv_tmp, inv_id)
115
mutter('new inventory_id is {%s}' % inv_id)
117
# We could also just sha hash the inv_tmp file
118
# however, in the case that branch.inventory_store.add()
119
# ever actually does anything special
120
inv_sha1 = branch.get_inventory_sha1(inv_id)
122
precursor = branch.last_patch()
124
precursor_sha1 = branch.get_revision_sha1(precursor)
126
precursor_sha1 = None
128
branch._write_inventory(work_inv)
130
if timestamp == None:
131
timestamp = time.time()
133
if committer == None:
134
committer = username()
137
timezone = local_time_offset()
139
mutter("building commit log message")
140
rev = Revision(timestamp=timestamp,
143
precursor = precursor,
144
precursor_sha1 = precursor_sha1,
147
inventory_sha1=inv_sha1,
150
rev_tmp = tempfile.TemporaryFile()
151
rev.write_xml(rev_tmp)
153
branch.revision_store.add(rev_tmp, rev_id)
154
mutter("new revision_id is {%s}" % rev_id)
156
## XXX: Everything up to here can simply be orphaned if we abort
157
## the commit; it will leave junk files behind but that doesn't
160
## TODO: Read back the just-generated changeset, and make sure it
161
## applies and recreates the right state.
163
## TODO: Also calculate and store the inventory SHA1
164
mutter("committing patch r%d" % (branch.revno() + 1))
166
branch.append_revision(rev_id)
169
note("commited r%d" % branch.revno())
175
def _gen_revision_id(when):
176
"""Return new revision-id."""
177
from binascii import hexlify
178
from osutils import rand_bytes, compact_date, user_email
180
s = '%s-%s-' % (user_email(), compact_date(when))
181
s += hexlify(rand_bytes(8))
185
def _gather_commit(branch, work_tree, work_inv, basis_inv, specific_files,
187
"""Build inventory preparatory to commit.
189
This adds any changed files into the text store, and sets their
190
test-id, sha and size in the returned inventory appropriately.
193
Modified to hold a list of files that have been deleted from
194
the working directory; these should be removed from the
197
from bzrlib.inventory import Inventory
198
from osutils import isdir, isfile, sha_string, quotefn, \
199
local_time_offset, username, kind_marker, is_inside_any
201
from branch import gen_file_id
202
from errors import BzrError
203
from revision import Revision
204
from bzrlib.trace import mutter, note
209
for path, entry in work_inv.iter_entries():
210
## TODO: Check that the file kind has not changed from the previous
211
## revision of this file (if any).
213
p = branch.abspath(path)
214
file_id = entry.file_id
215
mutter('commit prep file %s, id %r ' % (p, file_id))
217
if specific_files and not is_inside_any(specific_files, path):
218
if basis_inv.has_id(file_id):
219
# carry over with previous state
220
inv.add(basis_inv[file_id].copy())
222
# omit this from committed inventory
226
if not work_tree.has_id(file_id):
228
print('deleted %s%s' % (path, kind_marker(entry.kind)))
229
mutter(" file is missing, removing from inventory")
230
missing_ids.append(file_id)
233
# this is present in the new inventory; may be new, modified or
235
old_ie = basis_inv.has_id(file_id) and basis_inv[file_id]
241
old_kind = old_ie.kind
242
if old_kind != entry.kind:
243
raise BzrError("entry %r changed kind from %r to %r"
244
% (file_id, old_kind, entry.kind))
246
if entry.kind == 'directory':
248
raise BzrError("%s is entered as directory but not a directory"
250
elif entry.kind == 'file':
252
raise BzrError("%s is entered as file but is not a file" % quotefn(p))
254
new_sha1 = work_tree.get_file_sha1(file_id)
257
and old_ie.text_sha1 == new_sha1):
258
## assert content == basis.get_file(file_id).read()
259
entry.text_id = old_ie.text_id
260
entry.text_sha1 = new_sha1
261
entry.text_size = old_ie.text_size
262
mutter(' unchanged from previous text_id {%s}' %
265
content = file(p, 'rb').read()
267
# calculate the sha again, just in case the file contents
268
# changed since we updated the cache
269
entry.text_sha1 = sha_string(content)
270
entry.text_size = len(content)
272
entry.text_id = gen_file_id(entry.name)
273
branch.text_store.add(content, entry.text_id)
274
mutter(' stored with text_id {%s}' % entry.text_id)
277
marked = path + kind_marker(entry.kind)
279
print 'added', marked
280
elif old_ie == entry:
282
elif (old_ie.name == entry.name
283
and old_ie.parent_id == entry.parent_id):
284
print 'modified', marked
286
print 'renamed', marked
288
return missing_ids, inv
18
# The newly committed revision is going to have a shape corresponding
19
# to that of the working tree. Files that are not in the
20
# working tree and that were in the predecessor are reported as
21
# removed --- this can include files that were either removed from the
22
# inventory or deleted in the working tree. If they were only
23
# deleted from disk, they are removed from the working inventory.
25
# We then consider the remaining entries, which will be in the new
26
# version. Directory entries are simply copied across. File entries
27
# must be checked to see if a new version of the file should be
28
# recorded. For each parent revision tree, we check to see what
29
# version of the file was present. If the file was present in at
30
# least one tree, and if it was the same version in all the trees,
31
# then we can just refer to that version. Otherwise, a new version
32
# representing the merger of the file versions must be added.
34
# TODO: Update hashcache before and after - or does the WorkingTree
37
# TODO: Rather than mashing together the ancestry and storing it back,
38
# perhaps the weave should have single method which does it all in one
39
# go, avoiding a lot of redundant work.
41
# TODO: Perhaps give a warning if one of the revisions marked as
42
# merged is already in the ancestry, and then don't record it as a
45
# TODO: If the file is newly merged but unchanged from the version it
46
# merges from, then it should still be reported as newly added
47
# relative to the basis revision.
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
# the rest of the code; add a deprecation of the old name.
57
from cStringIO import StringIO
65
from bzrlib.branch import Branch
67
from bzrlib.errors import (BzrError, PointlessCommit,
71
from bzrlib.osutils import (kind_marker, isdir,isfile, is_inside_any,
72
is_inside_or_parent_of_any,
73
minimum_path_selection,
74
quotefn, sha_file, split_lines,
77
from bzrlib.testament import Testament
78
from bzrlib.trace import mutter, note, warning, is_quiet
79
from bzrlib.xml5 import serializer_v5
80
from bzrlib.inventory import InventoryEntry, make_entry
81
from bzrlib import symbol_versioning
82
from bzrlib.symbol_versioning import (deprecated_passed,
85
from bzrlib.workingtree import WorkingTree
86
from bzrlib.urlutils import unescape_for_display
90
class NullCommitReporter(object):
91
"""I report on progress of a commit."""
93
def started(self, revno, revid, location=None):
95
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
96
"to started.", DeprecationWarning,
100
def snapshot_change(self, change, path):
103
def completed(self, revno, rev_id):
106
def deleted(self, file_id):
109
def escaped(self, escape_count, message):
112
def missing(self, path):
115
def renamed(self, change, old_path, new_path):
118
def is_verbose(self):
122
class ReportCommitToLog(NullCommitReporter):
124
def _note(self, format, *args):
127
Subclasses may choose to override this method.
131
def snapshot_change(self, change, path):
132
if change == 'unchanged':
134
if change == 'added' and path == '':
136
self._note("%s %s", change, path)
138
def started(self, revno, rev_id, location=None):
139
if location is not None:
140
location = ' to: ' + unescape_for_display(location, 'utf-8')
142
# When started was added, location was only made optional by
143
# accident. Matt Nordhoff 20071129
144
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
145
"to started.", DeprecationWarning,
148
self._note('Committing%s', location)
150
def completed(self, revno, rev_id):
151
self._note('Committed revision %d.', revno)
153
def deleted(self, file_id):
154
self._note('deleted %s', file_id)
156
def escaped(self, escape_count, message):
157
self._note("replaced %d control characters in message", escape_count)
159
def missing(self, path):
160
self._note('missing %s', path)
162
def renamed(self, change, old_path, new_path):
163
self._note('%s %s => %s', change, old_path, new_path)
165
def is_verbose(self):
169
class Commit(object):
170
"""Task of committing a new revision.
172
This is a MethodObject: it accumulates state as the commit is
173
prepared, and then it is discarded. It doesn't represent
174
historical revisions, just the act of recording a new one.
177
Modified to hold a list of files that have been deleted from
178
the working directory; these should be removed from the
184
"""Create a Commit object.
186
:param reporter: the default reporter to use or None to decide later
188
self.reporter = reporter
198
allow_pointless=True,
206
message_callback=None,
208
"""Commit working copy as a new revision.
210
:param message: the commit message (it or message_callback is required)
212
:param timestamp: if not None, seconds-since-epoch for a
213
postdated/predated commit.
215
:param specific_files: If true, commit only those files.
217
:param rev_id: If set, use this as the new revision id.
218
Useful for test or import commands that need to tightly
219
control what revisions are assigned. If you duplicate
220
a revision id that exists elsewhere it is your own fault.
221
If null (default), a time/random revision id is generated.
223
:param allow_pointless: If true (default), commit even if nothing
224
has changed and no merges are recorded.
226
:param strict: If true, don't allow a commit if the working tree
227
contains unknown files.
229
:param revprops: Properties for new revision
230
:param local: Perform a local only commit.
231
:param reporter: the reporter to use or None for the default
232
:param verbose: if True and the reporter is not None, report everything
233
:param recursive: If set to 'down', commit in any subtrees that have
234
pending changes of any sort during this commit.
236
mutter('preparing to commit')
238
if working_tree is None:
239
raise BzrError("working_tree must be passed into commit().")
241
self.work_tree = working_tree
242
self.branch = self.work_tree.branch
243
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
244
if not self.branch.repository.supports_rich_root():
245
raise errors.RootNotRich()
246
if message_callback is None:
247
if message is not None:
248
if isinstance(message, str):
249
message = message.decode(bzrlib.user_encoding)
250
message_callback = lambda x: message
252
raise BzrError("The message or message_callback keyword"
253
" parameter is required for commit().")
255
self.bound_branch = None
256
self.any_entries_changed = False
257
self.any_entries_deleted = False
259
self.master_branch = None
260
self.master_locked = False
261
self.recursive = recursive
263
if specific_files is not None:
264
self.specific_files = sorted(
265
minimum_path_selection(specific_files))
267
self.specific_files = None
268
self.specific_file_ids = None
269
self.allow_pointless = allow_pointless
270
self.revprops = revprops
271
self.message_callback = message_callback
272
self.timestamp = timestamp
273
self.timezone = timezone
274
self.committer = committer
276
self.verbose = verbose
277
# accumulates an inventory delta to the basis entry, so we can make
278
# just the necessary updates to the workingtree's cached basis.
279
self._basis_delta = []
281
self.work_tree.lock_write()
282
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
283
self.basis_revid = self.work_tree.last_revision()
284
self.basis_tree = self.work_tree.basis_tree()
285
self.basis_tree.lock_read()
287
# Cannot commit with conflicts present.
288
if len(self.work_tree.conflicts()) > 0:
289
raise ConflictsInTree
291
# Setup the bound branch variables as needed.
292
self._check_bound_branch()
294
# Check that the working tree is up to date
295
old_revno, new_revno = self._check_out_of_date_tree()
297
# Complete configuration setup
298
if reporter is not None:
299
self.reporter = reporter
300
elif self.reporter is None:
301
self.reporter = self._select_reporter()
302
if self.config is None:
303
self.config = self.branch.get_config()
305
# If provided, ensure the specified files are versioned
306
if self.specific_files is not None:
307
# Note: This routine is being called because it raises
308
# PathNotVersionedError as a side effect of finding the IDs. We
309
# later use the ids we found as input to the working tree
310
# inventory iterator, so we only consider those ids rather than
311
# examining the whole tree again.
312
# XXX: Dont we have filter_unversioned to do this more
314
self.specific_file_ids = tree.find_ids_across_trees(
315
specific_files, [self.basis_tree, self.work_tree])
317
# Setup the progress bar. As the number of files that need to be
318
# committed in unknown, progress is reported as stages.
319
# We keep track of entries separately though and include that
320
# information in the progress bar during the relevant stages.
321
self.pb_stage_name = ""
322
self.pb_stage_count = 0
323
self.pb_stage_total = 5
324
if self.bound_branch:
325
self.pb_stage_total += 1
326
self.pb.show_pct = False
327
self.pb.show_spinner = False
328
self.pb.show_eta = False
329
self.pb.show_count = True
330
self.pb.show_bar = True
332
# After a merge, a selected file commit is not supported.
333
# See 'bzr help merge' for an explanation as to why.
334
self.basis_inv = self.basis_tree.inventory
335
self._gather_parents()
336
if len(self.parents) > 1 and self.specific_files:
337
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
339
# Collect the changes
340
self._set_progress_stage("Collecting changes",
341
entries_title="Directory")
342
self.builder = self.branch.get_commit_builder(self.parents,
343
self.config, timestamp, timezone, committer, revprops, rev_id)
346
# find the location being committed to
347
if self.bound_branch:
348
master_location = self.master_branch.base
350
master_location = self.branch.base
352
# report the start of the commit
353
self.reporter.started(new_revno, self.rev_id, master_location)
355
self._update_builder_with_changes()
356
self._report_and_accumulate_deletes()
357
self._check_pointless()
359
# TODO: Now the new inventory is known, check for conflicts.
360
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
361
# weave lines, because nothing should be recorded until it is known
362
# that commit will succeed.
363
self._set_progress_stage("Saving data locally")
364
self.builder.finish_inventory()
366
# Prompt the user for a commit message if none provided
367
message = message_callback(self)
368
assert isinstance(message, unicode), type(message)
369
self.message = message
370
self._escape_commit_message()
372
# Add revision data to the local branch
373
self.rev_id = self.builder.commit(self.message)
379
self._process_pre_hooks(old_revno, new_revno)
381
# Upload revision data to the master.
382
# this will propagate merged revisions too if needed.
383
if self.bound_branch:
384
if not self.master_branch.repository.has_same_location(
385
self.branch.repository):
386
self._set_progress_stage("Uploading data to master branch")
387
self.master_branch.repository.fetch(self.branch.repository,
388
revision_id=self.rev_id)
389
# now the master has the revision data
390
# 'commit' to the master first so a timeout here causes the
391
# local branch to be out of date
392
self.master_branch.set_last_revision_info(new_revno,
395
# and now do the commit locally.
396
self.branch.set_last_revision_info(new_revno, self.rev_id)
398
# Make the working tree up to date with the branch
399
self._set_progress_stage("Updating the working tree")
400
self.work_tree.update_basis_by_delta(self.rev_id,
402
self.reporter.completed(new_revno, self.rev_id)
403
self._process_post_hooks(old_revno, new_revno)
408
def _select_reporter(self):
409
"""Select the CommitReporter to use."""
411
return NullCommitReporter()
412
return ReportCommitToLog()
414
def _check_pointless(self):
415
if self.allow_pointless:
417
# A merge with no effect on files
418
if len(self.parents) > 1:
420
# TODO: we could simplify this by using self._basis_delta.
422
# The initial commit adds a root directory, but this in itself is not
423
# a worthwhile commit.
424
if (self.basis_revid == revision.NULL_REVISION and
425
len(self.builder.new_inventory) == 1):
426
raise PointlessCommit()
427
# If length == 1, then we only have the root entry. Which means
428
# that there is no real difference (only the root could be different)
429
# unless deletes occured, in which case the length is irrelevant.
430
if (self.any_entries_deleted or
431
(len(self.builder.new_inventory) != 1 and
432
self.any_entries_changed)):
434
raise PointlessCommit()
436
def _check_bound_branch(self):
437
"""Check to see if the local branch is bound.
439
If it is bound, then most of the commit will actually be
440
done using the remote branch as the target branch.
441
Only at the end will the local branch be updated.
443
if self.local and not self.branch.get_bound_location():
444
raise errors.LocalRequiresBoundBranch()
447
self.master_branch = self.branch.get_master_branch()
449
if not self.master_branch:
450
# make this branch the reference branch for out of date checks.
451
self.master_branch = self.branch
454
# If the master branch is bound, we must fail
455
master_bound_location = self.master_branch.get_bound_location()
456
if master_bound_location:
457
raise errors.CommitToDoubleBoundBranch(self.branch,
458
self.master_branch, master_bound_location)
460
# TODO: jam 20051230 We could automatically push local
461
# commits to the remote branch if they would fit.
462
# But for now, just require remote to be identical
465
# Make sure the local branch is identical to the master
466
master_info = self.master_branch.last_revision_info()
467
local_info = self.branch.last_revision_info()
468
if local_info != master_info:
469
raise errors.BoundBranchOutOfDate(self.branch,
472
# Now things are ready to change the master branch
474
self.bound_branch = self.branch
475
self.master_branch.lock_write()
476
self.master_locked = True
478
def _check_out_of_date_tree(self):
479
"""Check that the working tree is up to date.
481
:return: old_revision_number,new_revision_number tuple
484
first_tree_parent = self.work_tree.get_parent_ids()[0]
486
# if there are no parents, treat our parent as 'None'
487
# this is so that we still consider the master branch
488
# - in a checkout scenario the tree may have no
489
# parents but the branch may do.
490
first_tree_parent = bzrlib.revision.NULL_REVISION
491
old_revno, master_last = self.master_branch.last_revision_info()
492
if master_last != first_tree_parent:
493
if master_last != bzrlib.revision.NULL_REVISION:
494
raise errors.OutOfDateTree(self.work_tree)
495
if self.branch.repository.has_revision(first_tree_parent):
496
new_revno = old_revno + 1
498
# ghost parents never appear in revision history.
500
return old_revno,new_revno
502
def _process_pre_hooks(self, old_revno, new_revno):
503
"""Process any registered pre commit hooks."""
504
self._set_progress_stage("Running pre_commit hooks")
505
self._process_hooks("pre_commit", old_revno, new_revno)
507
def _process_post_hooks(self, old_revno, new_revno):
508
"""Process any registered post commit hooks."""
509
# Process the post commit hooks, if any
510
self._set_progress_stage("Running post_commit hooks")
511
# old style commit hooks - should be deprecated ? (obsoleted in
513
if self.config.post_commit() is not None:
514
hooks = self.config.post_commit().split(' ')
515
# this would be nicer with twisted.python.reflect.namedAny
517
result = eval(hook + '(branch, rev_id)',
518
{'branch':self.branch,
520
'rev_id':self.rev_id})
521
# process new style post commit hooks
522
self._process_hooks("post_commit", old_revno, new_revno)
524
def _process_hooks(self, hook_name, old_revno, new_revno):
525
if not Branch.hooks[hook_name]:
528
# new style commit hooks:
529
if not self.bound_branch:
530
hook_master = self.branch
533
hook_master = self.master_branch
534
hook_local = self.branch
535
# With bound branches, when the master is behind the local branch,
536
# the 'old_revno' and old_revid values here are incorrect.
537
# XXX: FIXME ^. RBC 20060206
539
old_revid = self.parents[0]
541
old_revid = bzrlib.revision.NULL_REVISION
543
if hook_name == "pre_commit":
544
future_tree = self.builder.revision_tree()
545
tree_delta = future_tree.changes_from(self.basis_tree,
548
for hook in Branch.hooks[hook_name]:
549
# show the running hook in the progress bar. As hooks may
550
# end up doing nothing (e.g. because they are not configured by
551
# the user) this is still showing progress, not showing overall
552
# actions - its up to each plugin to show a UI if it want's to
553
# (such as 'Emailing diff to foo@example.com').
554
self.pb_stage_name = "Running %s hooks [%s]" % \
555
(hook_name, Branch.hooks.get_hook_name(hook))
556
self._emit_progress()
557
if 'hooks' in debug.debug_flags:
558
mutter("Invoking commit hook: %r", hook)
559
if hook_name == "post_commit":
560
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
562
elif hook_name == "pre_commit":
563
hook(hook_local, hook_master,
564
old_revno, old_revid, new_revno, self.rev_id,
565
tree_delta, future_tree)
568
"""Cleanup any open locks, progress bars etc."""
569
cleanups = [self._cleanup_bound_branch,
570
self.basis_tree.unlock,
571
self.work_tree.unlock,
573
found_exception = None
574
for cleanup in cleanups:
577
# we want every cleanup to run no matter what.
578
# so we have a catchall here, but we will raise the
579
# last encountered exception up the stack: and
580
# typically this will be useful enough.
583
if found_exception is not None:
584
# don't do a plan raise, because the last exception may have been
585
# trashed, e is our sure-to-work exception even though it loses the
586
# full traceback. XXX: RBC 20060421 perhaps we could check the
587
# exc_info and if its the same one do a plain raise otherwise
588
# 'raise e' as we do now.
591
def _cleanup_bound_branch(self):
592
"""Executed at the end of a try/finally to cleanup a bound branch.
594
If the branch wasn't bound, this is a no-op.
595
If it was, it resents self.branch to the local branch, instead
598
if not self.bound_branch:
600
if self.master_locked:
601
self.master_branch.unlock()
603
def _escape_commit_message(self):
604
"""Replace xml-incompatible control characters."""
605
# FIXME: RBC 20060419 this should be done by the revision
606
# serialiser not by commit. Then we can also add an unescaper
607
# in the deserializer and start roundtripping revision messages
608
# precisely. See repository_implementations/test_repository.py
610
# Python strings can include characters that can't be
611
# represented in well-formed XML; escape characters that
612
# aren't listed in the XML specification
613
# (http://www.w3.org/TR/REC-xml/#NT-Char).
614
self.message, escape_count = re.subn(
615
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
616
lambda match: match.group(0).encode('unicode_escape'),
619
self.reporter.escaped(escape_count, self.message)
621
def _gather_parents(self):
622
"""Record the parents of a merge for merge detection."""
623
# TODO: Make sure that this list doesn't contain duplicate
624
# entries and the order is preserved when doing this.
625
self.parents = self.work_tree.get_parent_ids()
626
self.parent_invs = [self.basis_inv]
627
for revision in self.parents[1:]:
628
if self.branch.repository.has_revision(revision):
629
mutter('commit parent revision {%s}', revision)
630
inventory = self.branch.repository.get_inventory(revision)
631
self.parent_invs.append(inventory)
633
mutter('commit parent ghost revision {%s}', revision)
635
def _update_builder_with_changes(self):
636
"""Update the commit builder with the data about what has changed.
638
# Build the revision inventory.
640
# This starts by creating a new empty inventory. Depending on
641
# which files are selected for commit, and what is present in the
642
# current tree, the new inventory is populated. inventory entries
643
# which are candidates for modification have their revision set to
644
# None; inventory entries that are carried over untouched have their
645
# revision set to their prior value.
647
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
648
# results to create a new inventory at the same time, which results
649
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
652
specific_files = self.specific_files
653
mutter("Selecting files for commit with filter %s", specific_files)
655
# Build the new inventory
656
self._populate_from_inventory(specific_files)
658
# If specific files are selected, then all un-selected files must be
659
# recorded in their previous state. For more details, see
660
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
662
for path, old_ie in self.basis_inv.iter_entries():
663
if old_ie.file_id in self.builder.new_inventory:
664
# already added - skip.
666
if is_inside_any(specific_files, path):
667
# was inside the selected path, if not present it has been
670
if old_ie.kind == 'directory':
671
self._next_progress_entry()
672
# not in final inv yet, was not in the selected files, so is an
673
# entry to be preserved unaltered.
675
# Note: specific file commits after a merge are currently
676
# prohibited. This test is for sanity/safety in case it's
677
# required after that changes.
678
if len(self.parents) > 1:
680
delta, version_recorded = self.builder.record_entry_contents(
681
ie, self.parent_invs, path, self.basis_tree, None)
683
self.any_entries_changed = True
684
if delta: self._basis_delta.append(delta)
686
def _report_and_accumulate_deletes(self):
687
# XXX: Could the list of deleted paths and ids be instead taken from
688
# _populate_from_inventory?
689
deleted_ids = set(self.basis_inv._byid.keys()) - \
690
set(self.builder.new_inventory._byid.keys())
692
self.any_entries_deleted = True
693
deleted = [(self.basis_tree.id2path(file_id), file_id)
694
for file_id in deleted_ids]
696
# XXX: this is not quite directory-order sorting
697
for path, file_id in deleted:
698
self._basis_delta.append((path, None, file_id, None))
699
self.reporter.deleted(path)
701
def _populate_from_inventory(self, specific_files):
702
"""Populate the CommitBuilder by walking the working tree inventory."""
704
# raise an exception as soon as we find a single unknown.
705
for unknown in self.work_tree.unknowns():
706
raise StrictCommitFailed()
708
report_changes = self.reporter.is_verbose()
710
# A tree of paths that have been deleted. E.g. if foo/bar has been
711
# deleted, then we have {'foo':{'bar':{}}}
713
# XXX: Note that entries may have the wrong kind because the entry does
714
# not reflect the status on disk.
715
work_inv = self.work_tree.inventory
716
entries = work_inv.iter_entries_by_dir(
717
specific_file_ids=self.specific_file_ids, yield_parents=True)
718
for path, existing_ie in entries:
719
file_id = existing_ie.file_id
720
name = existing_ie.name
721
parent_id = existing_ie.parent_id
722
kind = existing_ie.kind
723
if kind == 'directory':
724
self._next_progress_entry()
725
# Skip files that have been deleted from the working tree.
726
# The deleted path ids are also recorded so they can be explicitly
729
path_segments = splitpath(path)
730
deleted_dict = deleted_paths
731
for segment in path_segments:
732
deleted_dict = deleted_dict.get(segment, None)
734
# We either took a path not present in the dict
735
# (deleted_dict was None), or we've reached an empty
736
# child dir in the dict, so are now a sub-path.
740
if deleted_dict is not None:
741
# the path has a deleted parent, do not add it.
743
content_summary = self.work_tree.path_content_summary(path)
744
# Note that when a filter of specific files is given, we must only
745
# skip/record deleted files matching that filter.
746
if not specific_files or is_inside_any(specific_files, path):
747
if content_summary[0] == 'missing':
748
if not deleted_paths:
749
# path won't have been split yet.
750
path_segments = splitpath(path)
751
deleted_dict = deleted_paths
752
for segment in path_segments:
753
deleted_dict = deleted_dict.setdefault(segment, {})
754
self.reporter.missing(path)
755
deleted_ids.append(file_id)
757
# TODO: have the builder do the nested commit just-in-time IF and
759
if content_summary[0] == 'tree-reference':
760
# enforce repository nested tree policy.
761
if (not self.work_tree.supports_tree_reference() or
762
# repository does not support it either.
763
not self.branch.repository._format.supports_tree_reference):
764
content_summary = ('directory',) + content_summary[1:]
765
kind = content_summary[0]
766
# TODO: specific_files filtering before nested tree processing
767
if kind == 'tree-reference':
768
if self.recursive == 'down':
769
nested_revision_id = self._commit_nested_tree(
771
content_summary = content_summary[:3] + (
774
content_summary = content_summary[:3] + (
775
self.work_tree.get_reference_revision(file_id),)
777
# Record an entry for this item
778
# Note: I don't particularly want to have the existing_ie
779
# parameter but the test suite currently (28-Jun-07) breaks
780
# without it thanks to a unicode normalisation issue. :-(
781
definitely_changed = kind != existing_ie.kind
782
self._record_entry(path, file_id, specific_files, kind, name,
783
parent_id, definitely_changed, existing_ie, report_changes,
786
# Unversion IDs that were found to be deleted
787
self.work_tree.unversion(deleted_ids)
789
def _commit_nested_tree(self, file_id, path):
790
"Commit a nested tree."
791
sub_tree = self.work_tree.get_nested_tree(file_id, path)
792
# FIXME: be more comprehensive here:
793
# this works when both trees are in --trees repository,
794
# but when both are bound to a different repository,
795
# it fails; a better way of approaching this is to
796
# finally implement the explicit-caches approach design
797
# a while back - RBC 20070306.
798
if sub_tree.branch.repository.has_same_location(
799
self.work_tree.branch.repository):
800
sub_tree.branch.repository = \
801
self.work_tree.branch.repository
803
return sub_tree.commit(message=None, revprops=self.revprops,
804
recursive=self.recursive,
805
message_callback=self.message_callback,
806
timestamp=self.timestamp, timezone=self.timezone,
807
committer=self.committer,
808
allow_pointless=self.allow_pointless,
809
strict=self.strict, verbose=self.verbose,
810
local=self.local, reporter=self.reporter)
811
except errors.PointlessCommit:
812
return self.work_tree.get_reference_revision(file_id)
814
def _record_entry(self, path, file_id, specific_files, kind, name,
815
parent_id, definitely_changed, existing_ie, report_changes,
817
"Record the new inventory entry for a path if any."
818
# mutter('check %s {%s}', path, file_id)
819
# mutter('%s selected for commit', path)
820
if definitely_changed or existing_ie is None:
821
ie = make_entry(kind, name, parent_id, file_id)
823
ie = existing_ie.copy()
825
delta, version_recorded = self.builder.record_entry_contents(ie,
826
self.parent_invs, path, self.work_tree, content_summary)
828
self._basis_delta.append(delta)
830
self.any_entries_changed = True
832
self._report_change(ie, path)
835
def _report_change(self, ie, path):
836
"""Report a change to the user.
838
The change that has occurred is described relative to the basis
841
if (self.basis_inv.has_id(ie.file_id)):
842
basis_ie = self.basis_inv[ie.file_id]
845
change = ie.describe_change(basis_ie, ie)
846
if change in (InventoryEntry.RENAMED,
847
InventoryEntry.MODIFIED_AND_RENAMED):
848
old_path = self.basis_inv.id2path(ie.file_id)
849
self.reporter.renamed(change, old_path, path)
851
self.reporter.snapshot_change(change, path)
853
def _set_progress_stage(self, name, entries_title=None):
854
"""Set the progress stage and emit an update to the progress bar."""
855
self.pb_stage_name = name
856
self.pb_stage_count += 1
857
self.pb_entries_title = entries_title
858
if entries_title is not None:
859
self.pb_entries_count = 0
860
self.pb_entries_total = '?'
861
self._emit_progress()
863
def _next_progress_entry(self):
864
"""Emit an update to the progress bar and increment the entry count."""
865
self.pb_entries_count += 1
866
self._emit_progress()
868
def _emit_progress(self):
869
if self.pb_entries_title:
870
if self.pb_entries_total == '?':
871
text = "%s [%s %d] - Stage" % (self.pb_stage_name,
872
self.pb_entries_title, self.pb_entries_count)
874
text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
875
self.pb_entries_title, self.pb_entries_count,
876
str(self.pb_entries_total))
878
text = "%s - Stage" % (self.pb_stage_name)
879
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
added
removed
bzrlib/commit.py
