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
# 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.
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18
# The newly committed revision is going to have a shape corresponding
19
# to that of the working tree. Files that are not in the
20
# working tree and that were in the predecessor are reported as
21
# removed --- this can include files that were either removed from the
22
# inventory or deleted in the working tree. If they were only
23
# deleted from disk, they are removed from the working inventory.
25
# We then consider the remaining entries, which will be in the new
26
# version. Directory entries are simply copied across. File entries
27
# must be checked to see if a new version of the file should be
28
# recorded. For each parent revision tree, we check to see what
29
# version of the file was present. If the file was present in at
30
# least one tree, and if it was the same version in all the trees,
31
# then we can just refer to that version. Otherwise, a new version
32
# representing the merger of the file versions must be added.
34
# TODO: Update hashcache before and after - or does the WorkingTree
37
# TODO: Rather than mashing together the ancestry and storing it back,
38
# perhaps the weave should have single method which does it all in one
39
# go, avoiding a lot of redundant work.
41
# TODO: Perhaps give a warning if one of the revisions marked as
42
# merged is already in the ancestry, and then don't record it as a
45
# TODO: If the file is newly merged but unchanged from the version it
46
# merges from, then it should still be reported as newly added
47
# relative to the basis revision.
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
# the rest of the code; add a deprecation of the old name.
57
from cStringIO import StringIO
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, path):
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 path == '' and change in ('added', 'modified'):
135
self._note("%s %s", change, path)
137
def started(self, revno, rev_id, location=None):
138
if location is not None:
139
location = ' to: ' + unescape_for_display(location, 'utf-8')
141
# When started was added, location was only made optional by
142
# accident. Matt Nordhoff 20071129
143
symbol_versioning.warn("As of bzr 1.0 you must pass a location "
144
"to started.", DeprecationWarning,
147
self._note('Committing%s', location)
149
def completed(self, revno, rev_id):
150
self._note('Committed revision %d.', revno)
152
def deleted(self, path):
153
self._note('deleted %s', path)
155
def escaped(self, escape_count, message):
156
self._note("replaced %d control characters in message", escape_count)
158
def missing(self, path):
159
self._note('missing %s', path)
161
def renamed(self, change, old_path, new_path):
162
self._note('%s %s => %s', change, old_path, new_path)
164
def is_verbose(self):
168
class Commit(object):
169
"""Task of committing a new revision.
171
This is a MethodObject: it accumulates state as the commit is
172
prepared, and then it is discarded. It doesn't represent
173
historical revisions, just the act of recording a new one.
176
Modified to hold a list of files that have been deleted from
177
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):
183
"""Create a Commit object.
185
:param reporter: the default reporter to use or None to decide later
187
self.reporter = reporter
197
allow_pointless=True,
205
message_callback=None,
208
possible_master_transports=None):
209
"""Commit working copy as a new revision.
211
:param message: the commit message (it or message_callback is required)
213
:param timestamp: if not None, seconds-since-epoch for a
214
postdated/predated commit.
216
:param specific_files: If true, commit only those files.
218
:param rev_id: If set, use this as the new revision id.
219
Useful for test or import commands that need to tightly
220
control what revisions are assigned. If you duplicate
221
a revision id that exists elsewhere it is your own fault.
222
If null (default), a time/random revision id is generated.
224
:param allow_pointless: If true (default), commit even if nothing
225
has changed and no merges are recorded.
227
:param strict: If true, don't allow a commit if the working tree
228
contains unknown files.
230
:param revprops: Properties for new revision
231
:param local: Perform a local only commit.
232
:param reporter: the reporter to use or None for the default
233
:param verbose: if True and the reporter is not None, report everything
234
:param recursive: If set to 'down', commit in any subtrees that have
235
pending changes of any sort during this commit.
236
:param exclude: None or a list of relative paths to exclude from the
237
commit. Pending changes to excluded files will be ignored by the
240
mutter('preparing to commit')
242
if working_tree is None:
243
raise BzrError("working_tree must be passed into commit().")
245
self.work_tree = working_tree
246
self.branch = self.work_tree.branch
247
if getattr(self.work_tree, 'requires_rich_root', lambda: False)():
248
if not self.branch.repository.supports_rich_root():
249
raise errors.RootNotRich()
250
if message_callback is None:
251
if message is not None:
252
if isinstance(message, str):
253
message = message.decode(get_user_encoding())
254
message_callback = lambda x: message
256
raise BzrError("The message or message_callback keyword"
257
" parameter is required for commit().")
259
self.bound_branch = None
260
self.any_entries_deleted = False
261
if exclude is not None:
262
self.exclude = sorted(
263
minimum_path_selection(exclude))
267
self.master_branch = None
268
self.master_locked = False
269
self.recursive = recursive
271
if specific_files is not None:
272
self.specific_files = sorted(
273
minimum_path_selection(specific_files))
275
self.specific_files = None
277
self.allow_pointless = allow_pointless
278
self.revprops = revprops
279
self.message_callback = message_callback
280
self.timestamp = timestamp
281
self.timezone = timezone
282
self.committer = committer
284
self.verbose = verbose
286
self.work_tree.lock_write()
287
self.parents = self.work_tree.get_parent_ids()
288
# We can use record_iter_changes IFF iter_changes is compatible with
289
# the command line parameters, and the repository has fast delta
290
# generation. See bug 347649.
291
self.use_record_iter_changes = (
292
not self.specific_files and
294
not self.branch.repository._format.supports_tree_reference and
295
(self.branch.repository._format.fast_deltas or
296
len(self.parents) < 2))
297
self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
298
self.basis_revid = self.work_tree.last_revision()
299
self.basis_tree = self.work_tree.basis_tree()
300
self.basis_tree.lock_read()
302
# Cannot commit with conflicts present.
303
if len(self.work_tree.conflicts()) > 0:
304
raise ConflictsInTree
306
# Setup the bound branch variables as needed.
307
self._check_bound_branch(possible_master_transports)
309
# Check that the working tree is up to date
310
old_revno, new_revno = self._check_out_of_date_tree()
312
# Complete configuration setup
313
if reporter is not None:
314
self.reporter = reporter
315
elif self.reporter is None:
316
self.reporter = self._select_reporter()
317
if self.config is None:
318
self.config = self.branch.get_config()
320
self._set_specific_file_ids()
322
# Setup the progress bar. As the number of files that need to be
323
# committed in unknown, progress is reported as stages.
324
# We keep track of entries separately though and include that
325
# information in the progress bar during the relevant stages.
326
self.pb_stage_name = ""
327
self.pb_stage_count = 0
328
self.pb_stage_total = 5
329
if self.bound_branch:
330
self.pb_stage_total += 1
331
self.pb.show_pct = False
332
self.pb.show_spinner = False
333
self.pb.show_eta = False
334
self.pb.show_count = True
335
self.pb.show_bar = True
337
self._gather_parents()
338
# After a merge, a selected file commit is not supported.
339
# See 'bzr help merge' for an explanation as to why.
340
if len(self.parents) > 1 and self.specific_files:
341
raise errors.CannotCommitSelectedFileMerge(self.specific_files)
342
# Excludes are a form of selected file commit.
343
if len(self.parents) > 1 and self.exclude:
344
raise errors.CannotCommitSelectedFileMerge(self.exclude)
346
# Collect the changes
347
self._set_progress_stage("Collecting changes", counter=True)
348
self.builder = self.branch.get_commit_builder(self.parents,
349
self.config, timestamp, timezone, committer, revprops, rev_id)
352
self.builder.will_record_deletes()
353
# find the location being committed to
354
if self.bound_branch:
355
master_location = self.master_branch.base
357
master_location = self.branch.base
359
# report the start of the commit
360
self.reporter.started(new_revno, self.rev_id, master_location)
362
self._update_builder_with_changes()
363
self._check_pointless()
365
# TODO: Now the new inventory is known, check for conflicts.
366
# ADHB 2006-08-08: If this is done, populate_new_inv should not add
367
# weave lines, because nothing should be recorded until it is known
368
# that commit will succeed.
369
self._set_progress_stage("Saving data locally")
370
self.builder.finish_inventory()
372
# Prompt the user for a commit message if none provided
373
message = message_callback(self)
374
self.message = message
375
self._escape_commit_message()
377
# Add revision data to the local branch
378
self.rev_id = self.builder.commit(self.message)
381
mutter("aborting commit write group because of exception:")
382
trace.log_exception_quietly()
383
note("aborting commit write group: %r" % (e,))
387
self._process_pre_hooks(old_revno, new_revno)
389
# Upload revision data to the master.
390
# this will propagate merged revisions too if needed.
391
if self.bound_branch:
392
self._set_progress_stage("Uploading data to master branch")
393
# 'commit' to the master first so a timeout here causes the
394
# local branch to be out of date
395
self.master_branch.import_last_revision_info(
396
self.branch.repository, new_revno, self.rev_id)
398
# and now do the commit locally.
399
self.branch.set_last_revision_info(new_revno, self.rev_id)
401
# Make the working tree up to date with the branch
402
self._set_progress_stage("Updating the working tree")
403
self.work_tree.update_basis_by_delta(self.rev_id,
404
self.builder.get_basis_delta())
405
self.reporter.completed(new_revno, self.rev_id)
406
self._process_post_hooks(old_revno, new_revno)
411
def _select_reporter(self):
412
"""Select the CommitReporter to use."""
414
return NullCommitReporter()
415
return ReportCommitToLog()
417
def _check_pointless(self):
418
if self.allow_pointless:
420
# A merge with no effect on files
421
if len(self.parents) > 1:
423
# TODO: we could simplify this by using self.builder.basis_delta.
425
# The initial commit adds a root directory, but this in itself is not
426
# a worthwhile commit.
427
if (self.basis_revid == revision.NULL_REVISION and
428
((self.builder.new_inventory is not None and
429
len(self.builder.new_inventory) == 1) or
430
len(self.builder._basis_delta) == 1)):
101
431
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
432
if self.builder.any_changes():
434
raise PointlessCommit()
436
def _check_bound_branch(self, possible_master_transports=None):
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(
448
possible_master_transports)
450
if not self.master_branch:
451
# make this branch the reference branch for out of date checks.
452
self.master_branch = self.branch
455
# If the master branch is bound, we must fail
456
master_bound_location = self.master_branch.get_bound_location()
457
if master_bound_location:
458
raise errors.CommitToDoubleBoundBranch(self.branch,
459
self.master_branch, master_bound_location)
461
# TODO: jam 20051230 We could automatically push local
462
# commits to the remote branch if they would fit.
463
# But for now, just require remote to be identical
466
# Make sure the local branch is identical to the master
467
master_info = self.master_branch.last_revision_info()
468
local_info = self.branch.last_revision_info()
469
if local_info != master_info:
470
raise errors.BoundBranchOutOfDate(self.branch,
473
# Now things are ready to change the master branch
475
self.bound_branch = self.branch
476
self.master_branch.lock_write()
477
self.master_locked = True
479
def _check_out_of_date_tree(self):
480
"""Check that the working tree is up to date.
482
:return: old_revision_number,new_revision_number tuple
485
first_tree_parent = self.work_tree.get_parent_ids()[0]
487
# if there are no parents, treat our parent as 'None'
488
# this is so that we still consider the master branch
489
# - in a checkout scenario the tree may have no
490
# parents but the branch may do.
491
first_tree_parent = bzrlib.revision.NULL_REVISION
492
old_revno, master_last = self.master_branch.last_revision_info()
493
if master_last != first_tree_parent:
494
if master_last != bzrlib.revision.NULL_REVISION:
495
raise errors.OutOfDateTree(self.work_tree)
496
if self.branch.repository.has_revision(first_tree_parent):
497
new_revno = old_revno + 1
499
# ghost parents never appear in revision history.
501
return old_revno,new_revno
503
def _process_pre_hooks(self, old_revno, new_revno):
504
"""Process any registered pre commit hooks."""
505
self._set_progress_stage("Running pre_commit hooks")
506
self._process_hooks("pre_commit", old_revno, new_revno)
508
def _process_post_hooks(self, old_revno, new_revno):
509
"""Process any registered post commit hooks."""
510
# Process the post commit hooks, if any
511
self._set_progress_stage("Running post_commit hooks")
512
# old style commit hooks - should be deprecated ? (obsoleted in
514
if self.config.post_commit() is not None:
515
hooks = self.config.post_commit().split(' ')
516
# this would be nicer with twisted.python.reflect.namedAny
518
result = eval(hook + '(branch, rev_id)',
519
{'branch':self.branch,
521
'rev_id':self.rev_id})
522
# process new style post commit hooks
523
self._process_hooks("post_commit", old_revno, new_revno)
525
def _process_hooks(self, hook_name, old_revno, new_revno):
526
if not Branch.hooks[hook_name]:
529
# new style commit hooks:
530
if not self.bound_branch:
531
hook_master = self.branch
534
hook_master = self.master_branch
535
hook_local = self.branch
536
# With bound branches, when the master is behind the local branch,
537
# the 'old_revno' and old_revid values here are incorrect.
538
# XXX: FIXME ^. RBC 20060206
540
old_revid = self.parents[0]
542
old_revid = bzrlib.revision.NULL_REVISION
544
if hook_name == "pre_commit":
545
future_tree = self.builder.revision_tree()
546
tree_delta = future_tree.changes_from(self.basis_tree,
549
for hook in Branch.hooks[hook_name]:
550
# show the running hook in the progress bar. As hooks may
551
# end up doing nothing (e.g. because they are not configured by
552
# the user) this is still showing progress, not showing overall
553
# actions - its up to each plugin to show a UI if it want's to
554
# (such as 'Emailing diff to foo@example.com').
555
self.pb_stage_name = "Running %s hooks [%s]" % \
556
(hook_name, Branch.hooks.get_hook_name(hook))
557
self._emit_progress()
558
if 'hooks' in debug.debug_flags:
559
mutter("Invoking commit hook: %r", hook)
560
if hook_name == "post_commit":
561
hook(hook_local, hook_master, old_revno, old_revid, new_revno,
563
elif hook_name == "pre_commit":
564
hook(hook_local, hook_master,
565
old_revno, old_revid, new_revno, self.rev_id,
566
tree_delta, future_tree)
569
"""Cleanup any open locks, progress bars etc."""
570
cleanups = [self._cleanup_bound_branch,
571
self.basis_tree.unlock,
572
self.work_tree.unlock,
574
found_exception = None
575
for cleanup in cleanups:
578
# we want every cleanup to run no matter what.
579
# so we have a catchall here, but we will raise the
580
# last encountered exception up the stack: and
581
# typically this will be useful enough.
584
if found_exception is not None:
585
# don't do a plan raise, because the last exception may have been
586
# trashed, e is our sure-to-work exception even though it loses the
587
# full traceback. XXX: RBC 20060421 perhaps we could check the
588
# exc_info and if its the same one do a plain raise otherwise
589
# 'raise e' as we do now.
592
def _cleanup_bound_branch(self):
593
"""Executed at the end of a try/finally to cleanup a bound branch.
595
If the branch wasn't bound, this is a no-op.
596
If it was, it resents self.branch to the local branch, instead
599
if not self.bound_branch:
601
if self.master_locked:
602
self.master_branch.unlock()
604
def _escape_commit_message(self):
605
"""Replace xml-incompatible control characters."""
606
# FIXME: RBC 20060419 this should be done by the revision
607
# serialiser not by commit. Then we can also add an unescaper
608
# in the deserializer and start roundtripping revision messages
609
# precisely. See repository_implementations/test_repository.py
611
# Python strings can include characters that can't be
612
# represented in well-formed XML; escape characters that
613
# aren't listed in the XML specification
614
# (http://www.w3.org/TR/REC-xml/#NT-Char).
615
self.message, escape_count = re.subn(
616
u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
617
lambda match: match.group(0).encode('unicode_escape'),
620
self.reporter.escaped(escape_count, self.message)
622
def _gather_parents(self):
623
"""Record the parents of a merge for merge detection."""
624
# TODO: Make sure that this list doesn't contain duplicate
625
# entries and the order is preserved when doing this.
626
if self.use_record_iter_changes:
628
self.basis_inv = self.basis_tree.inventory
629
self.parent_invs = [self.basis_inv]
630
for revision in self.parents[1:]:
631
if self.branch.repository.has_revision(revision):
632
mutter('commit parent revision {%s}', revision)
633
inventory = self.branch.repository.get_inventory(revision)
634
self.parent_invs.append(inventory)
636
mutter('commit parent ghost revision {%s}', revision)
638
def _update_builder_with_changes(self):
639
"""Update the commit builder with the data about what has changed.
641
exclude = self.exclude
642
specific_files = self.specific_files or []
643
mutter("Selecting files for commit with filter %s", specific_files)
646
if self.use_record_iter_changes:
647
iter_changes = self.work_tree.iter_changes(self.basis_tree)
648
iter_changes = self._filter_iter_changes(iter_changes)
649
for file_id, path, fs_hash in self.builder.record_iter_changes(
650
self.work_tree, self.basis_revid, iter_changes):
651
self.work_tree._observed_sha1(file_id, path, fs_hash)
653
# Build the new inventory
654
self._populate_from_inventory()
655
self._record_unselected()
656
self._report_and_accumulate_deletes()
658
def _filter_iter_changes(self, iter_changes):
659
"""Process iter_changes.
661
This method reports on the changes in iter_changes to the user, and
662
converts 'missing' entries in the iter_changes iterator to 'deleted'
663
entries. 'missing' entries have their
665
:param iter_changes: An iter_changes to process.
666
:return: A generator of changes.
668
reporter = self.reporter
669
report_changes = reporter.is_verbose()
671
for change in iter_changes:
673
old_path = change[1][0]
674
new_path = change[1][1]
675
versioned = change[3][1]
677
versioned = change[3][1]
678
if kind is None and versioned:
681
reporter.missing(new_path)
682
deleted_ids.append(change[0])
683
# Reset the new path (None) and new versioned flag (False)
684
change = (change[0], (change[1][0], None), change[2],
685
(change[3][0], False)) + change[4:]
686
elif kind == 'tree-reference':
687
if self.recursive == 'down':
688
self._commit_nested_tree(change[0], change[1][1])
689
if change[3][0] or change[3][1]:
693
reporter.deleted(old_path)
694
elif old_path is None:
695
reporter.snapshot_change('added', new_path)
696
elif old_path != new_path:
697
reporter.renamed('renamed', old_path, new_path)
700
self.work_tree.branch.repository._format.rich_root_data):
701
# Don't report on changes to '' in non rich root
703
reporter.snapshot_change('modified', new_path)
704
self._next_progress_entry()
705
# Unversion IDs that were found to be deleted
706
self.work_tree.unversion(deleted_ids)
708
def _record_unselected(self):
709
# If specific files are selected, then all un-selected files must be
710
# recorded in their previous state. For more details, see
711
# https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
712
if self.specific_files or self.exclude:
713
specific_files = self.specific_files or []
714
for path, old_ie in self.basis_inv.iter_entries():
715
if old_ie.file_id in self.builder.new_inventory:
716
# already added - skip.
718
if (is_inside_any(specific_files, path)
719
and not is_inside_any(self.exclude, path)):
720
# was inside the selected path, and not excluded - if not
721
# present it has been deleted so skip.
723
# From here down it was either not selected, or was excluded:
724
# We preserve the entry unaltered.
726
# Note: specific file commits after a merge are currently
727
# prohibited. This test is for sanity/safety in case it's
728
# required after that changes.
729
if len(self.parents) > 1:
731
self.builder.record_entry_contents(ie, self.parent_invs, path,
732
self.basis_tree, None)
734
def _report_and_accumulate_deletes(self):
735
if (isinstance(self.basis_inv, Inventory)
736
and isinstance(self.builder.new_inventory, Inventory)):
737
# the older Inventory classes provide a _byid dict, and building a
738
# set from the keys of this dict is substantially faster than even
739
# getting a set of ids from the inventory
741
# <lifeless> set(dict) is roughly the same speed as
742
# set(iter(dict)) and both are significantly slower than
744
deleted_ids = set(self.basis_inv._byid.keys()) - \
745
set(self.builder.new_inventory._byid.keys())
747
deleted_ids = set(self.basis_inv) - set(self.builder.new_inventory)
749
self.any_entries_deleted = True
750
deleted = [(self.basis_tree.id2path(file_id), file_id)
751
for file_id in deleted_ids]
753
# XXX: this is not quite directory-order sorting
754
for path, file_id in deleted:
755
self.builder.record_delete(path, file_id)
756
self.reporter.deleted(path)
758
def _check_strict(self):
759
# XXX: when we use iter_changes this would likely be faster if
760
# iter_changes would check for us (even in the presence of
763
# raise an exception as soon as we find a single unknown.
764
for unknown in self.work_tree.unknowns():
765
raise StrictCommitFailed()
767
def _populate_from_inventory(self):
768
"""Populate the CommitBuilder by walking the working tree inventory."""
769
# Build the revision inventory.
771
# This starts by creating a new empty inventory. Depending on
772
# which files are selected for commit, and what is present in the
773
# current tree, the new inventory is populated. inventory entries
774
# which are candidates for modification have their revision set to
775
# None; inventory entries that are carried over untouched have their
776
# revision set to their prior value.
778
# ESEPARATIONOFCONCERNS: this function is diffing and using the diff
779
# results to create a new inventory at the same time, which results
780
# in bugs like #46635. Any reason not to use/enhance Tree.changes_from?
783
specific_files = self.specific_files
784
exclude = self.exclude
785
report_changes = self.reporter.is_verbose()
787
# A tree of paths that have been deleted. E.g. if foo/bar has been
788
# deleted, then we have {'foo':{'bar':{}}}
790
# XXX: Note that entries may have the wrong kind because the entry does
791
# not reflect the status on disk.
792
work_inv = self.work_tree.inventory
793
# NB: entries will include entries within the excluded ids/paths
794
# because iter_entries_by_dir has no 'exclude' facility today.
795
entries = work_inv.iter_entries_by_dir(
796
specific_file_ids=self.specific_file_ids, yield_parents=True)
797
for path, existing_ie in entries:
798
file_id = existing_ie.file_id
799
name = existing_ie.name
800
parent_id = existing_ie.parent_id
801
kind = existing_ie.kind
802
# Skip files that have been deleted from the working tree.
803
# The deleted path ids are also recorded so they can be explicitly
806
path_segments = splitpath(path)
807
deleted_dict = deleted_paths
808
for segment in path_segments:
809
deleted_dict = deleted_dict.get(segment, None)
811
# We either took a path not present in the dict
812
# (deleted_dict was None), or we've reached an empty
813
# child dir in the dict, so are now a sub-path.
817
if deleted_dict is not None:
818
# the path has a deleted parent, do not add it.
820
if exclude and is_inside_any(exclude, path):
821
# Skip excluded paths. Excluded paths are processed by
822
# _update_builder_with_changes.
824
content_summary = self.work_tree.path_content_summary(path)
825
# Note that when a filter of specific files is given, we must only
826
# skip/record deleted files matching that filter.
827
if not specific_files or is_inside_any(specific_files, path):
828
if content_summary[0] == 'missing':
829
if not deleted_paths:
830
# path won't have been split yet.
831
path_segments = splitpath(path)
832
deleted_dict = deleted_paths
833
for segment in path_segments:
834
deleted_dict = deleted_dict.setdefault(segment, {})
835
self.reporter.missing(path)
836
self._next_progress_entry()
837
deleted_ids.append(file_id)
839
# TODO: have the builder do the nested commit just-in-time IF and
841
if content_summary[0] == 'tree-reference':
842
# enforce repository nested tree policy.
843
if (not self.work_tree.supports_tree_reference() or
844
# repository does not support it either.
845
not self.branch.repository._format.supports_tree_reference):
846
content_summary = ('directory',) + content_summary[1:]
847
kind = content_summary[0]
848
# TODO: specific_files filtering before nested tree processing
849
if kind == 'tree-reference':
850
if self.recursive == 'down':
851
nested_revision_id = self._commit_nested_tree(
853
content_summary = content_summary[:3] + (
856
content_summary = content_summary[:3] + (
857
self.work_tree.get_reference_revision(file_id),)
859
# Record an entry for this item
860
# Note: I don't particularly want to have the existing_ie
861
# parameter but the test suite currently (28-Jun-07) breaks
862
# without it thanks to a unicode normalisation issue. :-(
863
definitely_changed = kind != existing_ie.kind
864
self._record_entry(path, file_id, specific_files, kind, name,
865
parent_id, definitely_changed, existing_ie, report_changes,
868
# Unversion IDs that were found to be deleted
869
self.work_tree.unversion(deleted_ids)
871
def _commit_nested_tree(self, file_id, path):
872
"Commit a nested tree."
873
sub_tree = self.work_tree.get_nested_tree(file_id, path)
874
# FIXME: be more comprehensive here:
875
# this works when both trees are in --trees repository,
876
# but when both are bound to a different repository,
877
# it fails; a better way of approaching this is to
878
# finally implement the explicit-caches approach design
879
# a while back - RBC 20070306.
880
if sub_tree.branch.repository.has_same_location(
881
self.work_tree.branch.repository):
882
sub_tree.branch.repository = \
883
self.work_tree.branch.repository
885
return sub_tree.commit(message=None, revprops=self.revprops,
886
recursive=self.recursive,
887
message_callback=self.message_callback,
888
timestamp=self.timestamp, timezone=self.timezone,
889
committer=self.committer,
890
allow_pointless=self.allow_pointless,
891
strict=self.strict, verbose=self.verbose,
892
local=self.local, reporter=self.reporter)
893
except errors.PointlessCommit:
894
return self.work_tree.get_reference_revision(file_id)
896
def _record_entry(self, path, file_id, specific_files, kind, name,
897
parent_id, definitely_changed, existing_ie, report_changes,
899
"Record the new inventory entry for a path if any."
900
# mutter('check %s {%s}', path, file_id)
901
# mutter('%s selected for commit', path)
902
if definitely_changed or existing_ie is None:
903
ie = make_entry(kind, name, parent_id, file_id)
905
ie = existing_ie.copy()
907
# For carried over entries we don't care about the fs hash - the repo
908
# isn't generating a sha, so we're not saving computation time.
909
_, _, fs_hash = self.builder.record_entry_contents(
910
ie, self.parent_invs, path, self.work_tree, content_summary)
912
self._report_change(ie, path)
914
self.work_tree._observed_sha1(ie.file_id, path, fs_hash)
917
def _report_change(self, ie, path):
918
"""Report a change to the user.
920
The change that has occurred is described relative to the basis
923
if (self.basis_inv.has_id(ie.file_id)):
924
basis_ie = self.basis_inv[ie.file_id]
927
change = ie.describe_change(basis_ie, ie)
928
if change in (InventoryEntry.RENAMED,
929
InventoryEntry.MODIFIED_AND_RENAMED):
930
old_path = self.basis_inv.id2path(ie.file_id)
931
self.reporter.renamed(change, old_path, path)
932
self._next_progress_entry()
934
if change == 'unchanged':
936
self.reporter.snapshot_change(change, path)
937
self._next_progress_entry()
939
def _set_progress_stage(self, name, counter=False):
940
"""Set the progress stage and emit an update to the progress bar."""
941
self.pb_stage_name = name
942
self.pb_stage_count += 1
944
self.pb_entries_count = 0
946
self.pb_entries_count = None
947
self._emit_progress()
949
def _next_progress_entry(self):
950
"""Emit an update to the progress bar and increment the entry count."""
951
self.pb_entries_count += 1
952
self._emit_progress()
954
def _emit_progress(self):
955
if self.pb_entries_count is not None:
956
text = "%s [%d] - Stage" % (self.pb_stage_name,
957
self.pb_entries_count)
959
text = "%s - Stage" % (self.pb_stage_name, )
960
self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
962
def _set_specific_file_ids(self):
963
"""populate self.specific_file_ids if we will use it."""
964
if not self.use_record_iter_changes:
965
# If provided, ensure the specified files are versioned
966
if self.specific_files is not None:
967
# Note: This routine is being called because it raises
968
# PathNotVersionedError as a side effect of finding the IDs. We
969
# later use the ids we found as input to the working tree
970
# inventory iterator, so we only consider those ids rather than
971
# examining the whole tree again.
972
# XXX: Dont we have filter_unversioned to do this more
974
self.specific_file_ids = tree.find_ids_across_trees(
975
self.specific_files, [self.basis_tree, self.work_tree])
977
self.specific_file_ids = None
added
removed
bzrlib/commit.py
