/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar

« back to all changes in this revision

Viewing changes to bzrlib/commit.py

  • Committer: Sabin Iacob
  • Date: 2009-03-23 14:59:43 UTC
  • mto: (4189.1.1 integration)
  • mto: This revision was merged to the branch mainline in revision 4193.
  • Revision ID: iacobs@m0n5t3r.info-20090323145943-3s3p1px5q1rkh2e5
update FSF mailing address

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005-2011 Canonical Ltd
 
1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
2
2
#
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
49
49
# TODO: Change the parameter 'rev_id' to 'revision_id' to be consistent with
50
50
# the rest of the code; add a deprecation of the old name.
51
51
 
 
52
import os
 
53
import re
 
54
import sys
 
55
import time
 
56
 
 
57
from cStringIO import StringIO
 
58
 
52
59
from bzrlib import (
53
60
    debug,
54
61
    errors,
55
62
    revision,
56
63
    trace,
57
64
    tree,
58
 
    ui,
59
65
    )
60
66
from bzrlib.branch import Branch
61
 
from bzrlib.cleanup import OperationWithCleanups
62
67
import bzrlib.config
63
68
from bzrlib.errors import (BzrError, PointlessCommit,
64
69
                           ConflictsInTree,
65
70
                           StrictCommitFailed
66
71
                           )
67
72
from bzrlib.osutils import (get_user_encoding,
68
 
                            is_inside_any,
 
73
                            kind_marker, isdir,isfile, is_inside_any,
 
74
                            is_inside_or_parent_of_any,
69
75
                            minimum_path_selection,
 
76
                            quotefn, sha_file, split_lines,
70
77
                            splitpath,
71
78
                            )
72
 
from bzrlib.trace import mutter, note, is_quiet
 
79
from bzrlib.testament import Testament
 
80
from bzrlib.trace import mutter, note, warning, is_quiet
73
81
from bzrlib.inventory import Inventory, InventoryEntry, make_entry
74
82
from bzrlib import symbol_versioning
 
83
from bzrlib.symbol_versioning import (deprecated_passed,
 
84
        deprecated_function,
 
85
        DEPRECATED_PARAMETER)
 
86
from bzrlib.workingtree import WorkingTree
75
87
from bzrlib.urlutils import unescape_for_display
 
88
import bzrlib.ui
76
89
 
77
90
 
78
91
class NullCommitReporter(object):
91
104
    def completed(self, revno, rev_id):
92
105
        pass
93
106
 
94
 
    def deleted(self, path):
 
107
    def deleted(self, file_id):
 
108
        pass
 
109
 
 
110
    def escaped(self, escape_count, message):
95
111
        pass
96
112
 
97
113
    def missing(self, path):
114
130
        note(format, *args)
115
131
 
116
132
    def snapshot_change(self, change, path):
117
 
        if path == '' and change in ('added', 'modified'):
 
133
        if change == 'unchanged':
 
134
            return
 
135
        if change == 'added' and path == '':
118
136
            return
119
137
        self._note("%s %s", change, path)
120
138
 
132
150
 
133
151
    def completed(self, revno, rev_id):
134
152
        self._note('Committed revision %d.', revno)
135
 
        # self._note goes to the console too; so while we want to log the
136
 
        # rev_id, we can't trivially only log it. (See bug 526425). Long
137
 
        # term we should rearrange the reporting structure, but for now
138
 
        # we just mutter seperately. We mutter the revid and revno together
139
 
        # so that concurrent bzr invocations won't lead to confusion.
140
 
        mutter('Committed revid %s as revno %d.', rev_id, revno)
141
 
 
142
 
    def deleted(self, path):
143
 
        self._note('deleted %s', path)
 
153
 
 
154
    def deleted(self, file_id):
 
155
        self._note('deleted %s', file_id)
 
156
 
 
157
    def escaped(self, escape_count, message):
 
158
        self._note("replaced %d control characters in message", escape_count)
144
159
 
145
160
    def missing(self, path):
146
161
        self._note('missing %s', path)
174
189
        self.reporter = reporter
175
190
        self.config = config
176
191
 
177
 
    @staticmethod
178
 
    def update_revprops(revprops, branch, authors=None, author=None,
179
 
                        local=False, possible_master_transports=None):
180
 
        if revprops is None:
181
 
            revprops = {}
182
 
        if possible_master_transports is None:
183
 
            possible_master_transports = []
184
 
        if not 'branch-nick' in revprops:
185
 
            revprops['branch-nick'] = branch._get_nick(
186
 
                local,
187
 
                possible_master_transports)
188
 
        if authors is not None:
189
 
            if author is not None:
190
 
                raise AssertionError('Specifying both author and authors '
191
 
                        'is not allowed. Specify just authors instead')
192
 
            if 'author' in revprops or 'authors' in revprops:
193
 
                # XXX: maybe we should just accept one of them?
194
 
                raise AssertionError('author property given twice')
195
 
            if authors:
196
 
                for individual in authors:
197
 
                    if '\n' in individual:
198
 
                        raise AssertionError('\\n is not a valid character '
199
 
                                'in an author identity')
200
 
                revprops['authors'] = '\n'.join(authors)
201
 
        if author is not None:
202
 
            symbol_versioning.warn('The parameter author was deprecated'
203
 
                   ' in version 1.13. Use authors instead',
204
 
                   DeprecationWarning)
205
 
            if 'author' in revprops or 'authors' in revprops:
206
 
                # XXX: maybe we should just accept one of them?
207
 
                raise AssertionError('author property given twice')
208
 
            if '\n' in author:
209
 
                raise AssertionError('\\n is not a valid character '
210
 
                        'in an author identity')
211
 
            revprops['authors'] = author
212
 
        return revprops
213
 
 
214
192
    def commit(self,
215
193
               message=None,
216
194
               timestamp=None,
229
207
               message_callback=None,
230
208
               recursive='down',
231
209
               exclude=None,
232
 
               possible_master_transports=None,
233
 
               lossy=False):
 
210
               possible_master_transports=None):
234
211
        """Commit working copy as a new revision.
235
212
 
236
213
        :param message: the commit message (it or message_callback is required)
237
 
        :param message_callback: A callback: message = message_callback(cmt_obj)
238
214
 
239
215
        :param timestamp: if not None, seconds-since-epoch for a
240
216
            postdated/predated commit.
241
217
 
242
 
        :param specific_files: If not None, commit only those files. An empty
243
 
            list means 'commit no files'.
 
218
        :param specific_files: If true, commit only those files.
244
219
 
245
220
        :param rev_id: If set, use this as the new revision id.
246
221
            Useful for test or import commands that need to tightly
263
238
        :param exclude: None or a list of relative paths to exclude from the
264
239
            commit. Pending changes to excluded files will be ignored by the
265
240
            commit.
266
 
        :param lossy: When committing to a foreign VCS, ignore any
267
 
            data that can not be natively represented.
268
241
        """
269
 
        operation = OperationWithCleanups(self._commit)
270
 
        self.revprops = revprops or {}
271
 
        # XXX: Can be set on __init__ or passed in - this is a bit ugly.
272
 
        self.config = config or self.config
273
 
        return operation.run(
274
 
               message=message,
275
 
               timestamp=timestamp,
276
 
               timezone=timezone,
277
 
               committer=committer,
278
 
               specific_files=specific_files,
279
 
               rev_id=rev_id,
280
 
               allow_pointless=allow_pointless,
281
 
               strict=strict,
282
 
               verbose=verbose,
283
 
               working_tree=working_tree,
284
 
               local=local,
285
 
               reporter=reporter,
286
 
               message_callback=message_callback,
287
 
               recursive=recursive,
288
 
               exclude=exclude,
289
 
               possible_master_transports=possible_master_transports,
290
 
               lossy=lossy)
291
 
 
292
 
    def _commit(self, operation, message, timestamp, timezone, committer,
293
 
            specific_files, rev_id, allow_pointless, strict, verbose,
294
 
            working_tree, local, reporter, message_callback, recursive,
295
 
            exclude, possible_master_transports, lossy):
296
242
        mutter('preparing to commit')
297
243
 
298
244
        if working_tree is None:
321
267
            self.exclude = []
322
268
        self.local = local
323
269
        self.master_branch = None
 
270
        self.master_locked = False
324
271
        self.recursive = recursive
325
272
        self.rev_id = None
326
 
        # self.specific_files is None to indicate no filter, or any iterable to
327
 
        # indicate a filter - [] means no files at all, as per iter_changes.
328
273
        if specific_files is not None:
329
274
            self.specific_files = sorted(
330
275
                minimum_path_selection(specific_files))
331
276
        else:
332
277
            self.specific_files = None
333
 
 
 
278
        self.specific_file_ids = None
334
279
        self.allow_pointless = allow_pointless
 
280
        self.revprops = revprops
335
281
        self.message_callback = message_callback
336
282
        self.timestamp = timestamp
337
283
        self.timezone = timezone
340
286
        self.verbose = verbose
341
287
 
342
288
        self.work_tree.lock_write()
343
 
        operation.add_cleanup(self.work_tree.unlock)
344
 
        self.parents = self.work_tree.get_parent_ids()
345
 
        # We can use record_iter_changes IFF iter_changes is compatible with
346
 
        # the command line parameters, and the repository has fast delta
347
 
        # generation. See bug 347649.
348
 
        self.use_record_iter_changes = (
349
 
            not self.exclude and 
350
 
            not self.branch.repository._format.supports_tree_reference and
351
 
            (self.branch.repository._format.fast_deltas or
352
 
             len(self.parents) < 2))
353
 
        self.pb = ui.ui_factory.nested_progress_bar()
354
 
        operation.add_cleanup(self.pb.finished)
 
289
        self.pb = bzrlib.ui.ui_factory.nested_progress_bar()
355
290
        self.basis_revid = self.work_tree.last_revision()
356
291
        self.basis_tree = self.work_tree.basis_tree()
357
292
        self.basis_tree.lock_read()
358
 
        operation.add_cleanup(self.basis_tree.unlock)
359
 
        # Cannot commit with conflicts present.
360
 
        if len(self.work_tree.conflicts()) > 0:
361
 
            raise ConflictsInTree
362
 
 
363
 
        # Setup the bound branch variables as needed.
364
 
        self._check_bound_branch(operation, possible_master_transports)
365
 
 
366
 
        # Check that the working tree is up to date
367
 
        old_revno, new_revno = self._check_out_of_date_tree()
368
 
 
369
 
        # Complete configuration setup
370
 
        if reporter is not None:
371
 
            self.reporter = reporter
372
 
        elif self.reporter is None:
373
 
            self.reporter = self._select_reporter()
374
 
        if self.config is None:
375
 
            self.config = self.branch.get_config()
376
 
 
377
 
        self._set_specific_file_ids()
378
 
 
379
 
        # Setup the progress bar. As the number of files that need to be
380
 
        # committed in unknown, progress is reported as stages.
381
 
        # We keep track of entries separately though and include that
382
 
        # information in the progress bar during the relevant stages.
383
 
        self.pb_stage_name = ""
384
 
        self.pb_stage_count = 0
385
 
        self.pb_stage_total = 5
386
 
        if self.bound_branch:
387
 
            # 2 extra stages: "Uploading data to master branch" and "Merging
388
 
            # tags to master branch"
389
 
            self.pb_stage_total += 2
390
 
        self.pb.show_pct = False
391
 
        self.pb.show_spinner = False
392
 
        self.pb.show_eta = False
393
 
        self.pb.show_count = True
394
 
        self.pb.show_bar = True
395
 
 
396
 
        self._gather_parents()
397
 
        # After a merge, a selected file commit is not supported.
398
 
        # See 'bzr help merge' for an explanation as to why.
399
 
        if len(self.parents) > 1 and self.specific_files is not None:
400
 
            raise errors.CannotCommitSelectedFileMerge(self.specific_files)
401
 
        # Excludes are a form of selected file commit.
402
 
        if len(self.parents) > 1 and self.exclude:
403
 
            raise errors.CannotCommitSelectedFileMerge(self.exclude)
404
 
 
405
 
        # Collect the changes
406
 
        self._set_progress_stage("Collecting changes", counter=True)
407
 
        self._lossy = lossy
408
 
        self.builder = self.branch.get_commit_builder(self.parents,
409
 
            self.config, timestamp, timezone, committer, self.revprops,
410
 
            rev_id, lossy=lossy)
411
 
        if not self.builder.supports_record_entry_contents and self.exclude:
412
 
            self.builder.abort()
413
 
            raise errors.ExcludesUnsupported(self.branch.repository)
414
 
 
415
293
        try:
416
 
            self.builder.will_record_deletes()
417
 
            # find the location being committed to
418
 
            if self.bound_branch:
419
 
                master_location = self.master_branch.base
420
 
            else:
421
 
                master_location = self.branch.base
422
 
 
423
 
            # report the start of the commit
424
 
            self.reporter.started(new_revno, self.rev_id, master_location)
425
 
 
426
 
            self._update_builder_with_changes()
427
 
            self._check_pointless()
428
 
 
429
 
            # TODO: Now the new inventory is known, check for conflicts.
430
 
            # ADHB 2006-08-08: If this is done, populate_new_inv should not add
431
 
            # weave lines, because nothing should be recorded until it is known
432
 
            # that commit will succeed.
433
 
            self._set_progress_stage("Saving data locally")
434
 
            self.builder.finish_inventory()
435
 
 
436
 
            # Prompt the user for a commit message if none provided
437
 
            message = message_callback(self)
438
 
            self.message = message
439
 
 
440
 
            # Add revision data to the local branch
441
 
            self.rev_id = self.builder.commit(self.message)
442
 
 
443
 
        except Exception, e:
444
 
            mutter("aborting commit write group because of exception:")
445
 
            trace.log_exception_quietly()
446
 
            note("aborting commit write group: %r" % (e,))
447
 
            self.builder.abort()
448
 
            raise
449
 
 
450
 
        self._process_pre_hooks(old_revno, new_revno)
451
 
 
452
 
        # Upload revision data to the master.
453
 
        # this will propagate merged revisions too if needed.
454
 
        if self.bound_branch:
455
 
            self._set_progress_stage("Uploading data to master branch")
456
 
            # 'commit' to the master first so a timeout here causes the
457
 
            # local branch to be out of date
458
 
            (new_revno, self.rev_id) = self.master_branch.import_last_revision_info_and_tags(
459
 
                self.branch, new_revno, self.rev_id, lossy=lossy)
460
 
            if lossy:
461
 
                self.branch.fetch(self.master_branch, self.rev_id)
462
 
 
463
 
        # and now do the commit locally.
464
 
        self.branch.set_last_revision_info(new_revno, self.rev_id)
465
 
 
466
 
        # Merge local tags to remote
467
 
        if self.bound_branch:
468
 
            self._set_progress_stage("Merging tags to master branch")
469
 
            tag_conflicts = self.branch.tags.merge_to(self.master_branch.tags)
470
 
            if tag_conflicts:
471
 
                warning_lines = ['    ' + name for name, _, _ in tag_conflicts]
472
 
                note("Conflicting tags in bound branch:\n" +
473
 
                    "\n".join(warning_lines))
474
 
 
475
 
        # Make the working tree be up to date with the branch. This
476
 
        # includes automatic changes scheduled to be made to the tree, such
477
 
        # as updating its basis and unversioning paths that were missing.
478
 
        self.work_tree.unversion(self.deleted_ids)
479
 
        self._set_progress_stage("Updating the working tree")
480
 
        self.work_tree.update_basis_by_delta(self.rev_id,
481
 
             self.builder.get_basis_delta())
482
 
        self.reporter.completed(new_revno, self.rev_id)
483
 
        self._process_post_hooks(old_revno, new_revno)
 
294
            # Cannot commit with conflicts present.
 
295
            if len(self.work_tree.conflicts()) > 0:
 
296
                raise ConflictsInTree
 
297
 
 
298
            # Setup the bound branch variables as needed.
 
299
            self._check_bound_branch(possible_master_transports)
 
300
 
 
301
            # Check that the working tree is up to date
 
302
            old_revno, new_revno = self._check_out_of_date_tree()
 
303
 
 
304
            # Complete configuration setup
 
305
            if reporter is not None:
 
306
                self.reporter = reporter
 
307
            elif self.reporter is None:
 
308
                self.reporter = self._select_reporter()
 
309
            if self.config is None:
 
310
                self.config = self.branch.get_config()
 
311
 
 
312
            # If provided, ensure the specified files are versioned
 
313
            if self.specific_files is not None:
 
314
                # Note: This routine is being called because it raises
 
315
                # PathNotVersionedError as a side effect of finding the IDs. We
 
316
                # later use the ids we found as input to the working tree
 
317
                # inventory iterator, so we only consider those ids rather than
 
318
                # examining the whole tree again.
 
319
                # XXX: Dont we have filter_unversioned to do this more
 
320
                # cheaply?
 
321
                self.specific_file_ids = tree.find_ids_across_trees(
 
322
                    specific_files, [self.basis_tree, self.work_tree])
 
323
 
 
324
            # Setup the progress bar. As the number of files that need to be
 
325
            # committed in unknown, progress is reported as stages.
 
326
            # We keep track of entries separately though and include that
 
327
            # information in the progress bar during the relevant stages.
 
328
            self.pb_stage_name = ""
 
329
            self.pb_stage_count = 0
 
330
            self.pb_stage_total = 5
 
331
            if self.bound_branch:
 
332
                self.pb_stage_total += 1
 
333
            self.pb.show_pct = False
 
334
            self.pb.show_spinner = False
 
335
            self.pb.show_eta = False
 
336
            self.pb.show_count = True
 
337
            self.pb.show_bar = True
 
338
 
 
339
            self.basis_inv = self.basis_tree.inventory
 
340
            self._gather_parents()
 
341
            # After a merge, a selected file commit is not supported.
 
342
            # See 'bzr help merge' for an explanation as to why.
 
343
            if len(self.parents) > 1 and self.specific_files:
 
344
                raise errors.CannotCommitSelectedFileMerge(self.specific_files)
 
345
            # Excludes are a form of selected file commit.
 
346
            if len(self.parents) > 1 and self.exclude:
 
347
                raise errors.CannotCommitSelectedFileMerge(self.exclude)
 
348
 
 
349
            # Collect the changes
 
350
            self._set_progress_stage("Collecting changes",
 
351
                    entries_title="Directory")
 
352
            self.builder = self.branch.get_commit_builder(self.parents,
 
353
                self.config, timestamp, timezone, committer, revprops, rev_id)
 
354
 
 
355
            try:
 
356
                self.builder.will_record_deletes()
 
357
                # find the location being committed to
 
358
                if self.bound_branch:
 
359
                    master_location = self.master_branch.base
 
360
                else:
 
361
                    master_location = self.branch.base
 
362
 
 
363
                # report the start of the commit
 
364
                self.reporter.started(new_revno, self.rev_id, master_location)
 
365
 
 
366
                self._update_builder_with_changes()
 
367
                self._report_and_accumulate_deletes()
 
368
                self._check_pointless()
 
369
 
 
370
                # TODO: Now the new inventory is known, check for conflicts.
 
371
                # ADHB 2006-08-08: If this is done, populate_new_inv should not add
 
372
                # weave lines, because nothing should be recorded until it is known
 
373
                # that commit will succeed.
 
374
                self._set_progress_stage("Saving data locally")
 
375
                self.builder.finish_inventory()
 
376
 
 
377
                # Prompt the user for a commit message if none provided
 
378
                message = message_callback(self)
 
379
                self.message = message
 
380
                self._escape_commit_message()
 
381
 
 
382
                # Add revision data to the local branch
 
383
                self.rev_id = self.builder.commit(self.message)
 
384
 
 
385
            except Exception, e:
 
386
                mutter("aborting commit write group because of exception:")
 
387
                trace.log_exception_quietly()
 
388
                note("aborting commit write group: %r" % (e,))
 
389
                self.builder.abort()
 
390
                raise
 
391
 
 
392
            self._process_pre_hooks(old_revno, new_revno)
 
393
 
 
394
            # Upload revision data to the master.
 
395
            # this will propagate merged revisions too if needed.
 
396
            if self.bound_branch:
 
397
                self._set_progress_stage("Uploading data to master branch")
 
398
                # 'commit' to the master first so a timeout here causes the
 
399
                # local branch to be out of date
 
400
                self.master_branch.import_last_revision_info(
 
401
                    self.branch.repository, new_revno, self.rev_id)
 
402
 
 
403
            # and now do the commit locally.
 
404
            self.branch.set_last_revision_info(new_revno, self.rev_id)
 
405
 
 
406
            # Make the working tree up to date with the branch
 
407
            self._set_progress_stage("Updating the working tree")
 
408
            self.work_tree.update_basis_by_delta(self.rev_id,
 
409
                 self.builder.get_basis_delta())
 
410
            self.reporter.completed(new_revno, self.rev_id)
 
411
            self._process_post_hooks(old_revno, new_revno)
 
412
        finally:
 
413
            self._cleanup()
484
414
        return self.rev_id
485
415
 
486
416
    def _select_reporter(self):
500
430
        # The initial commit adds a root directory, but this in itself is not
501
431
        # a worthwhile commit.
502
432
        if (self.basis_revid == revision.NULL_REVISION and
503
 
            ((self.builder.new_inventory is not None and
504
 
             len(self.builder.new_inventory) == 1) or
505
 
            len(self.builder._basis_delta) == 1)):
 
433
            len(self.builder.new_inventory) == 1):
506
434
            raise PointlessCommit()
507
 
        if self.builder.any_changes():
 
435
        # If length == 1, then we only have the root entry. Which means
 
436
        # that there is no real difference (only the root could be different)
 
437
        # unless deletes occured, in which case the length is irrelevant.
 
438
        if (self.any_entries_deleted or
 
439
            (len(self.builder.new_inventory) != 1 and
 
440
             self.builder.any_changes())):
508
441
            return
509
442
        raise PointlessCommit()
510
443
 
511
 
    def _check_bound_branch(self, operation, possible_master_transports=None):
 
444
    def _check_bound_branch(self, possible_master_transports=None):
512
445
        """Check to see if the local branch is bound.
513
446
 
514
447
        If it is bound, then most of the commit will actually be
549
482
        # so grab the lock
550
483
        self.bound_branch = self.branch
551
484
        self.master_branch.lock_write()
552
 
        operation.add_cleanup(self.master_branch.unlock)
 
485
        self.master_locked = True
553
486
 
554
487
    def _check_out_of_date_tree(self):
555
488
        """Check that the working tree is up to date.
640
573
                     old_revno, old_revid, new_revno, self.rev_id,
641
574
                     tree_delta, future_tree)
642
575
 
 
576
    def _cleanup(self):
 
577
        """Cleanup any open locks, progress bars etc."""
 
578
        cleanups = [self._cleanup_bound_branch,
 
579
                    self.basis_tree.unlock,
 
580
                    self.work_tree.unlock,
 
581
                    self.pb.finished]
 
582
        found_exception = None
 
583
        for cleanup in cleanups:
 
584
            try:
 
585
                cleanup()
 
586
            # we want every cleanup to run no matter what.
 
587
            # so we have a catchall here, but we will raise the
 
588
            # last encountered exception up the stack: and
 
589
            # typically this will be useful enough.
 
590
            except Exception, e:
 
591
                found_exception = e
 
592
        if found_exception is not None:
 
593
            # don't do a plan raise, because the last exception may have been
 
594
            # trashed, e is our sure-to-work exception even though it loses the
 
595
            # full traceback. XXX: RBC 20060421 perhaps we could check the
 
596
            # exc_info and if its the same one do a plain raise otherwise
 
597
            # 'raise e' as we do now.
 
598
            raise e
 
599
 
 
600
    def _cleanup_bound_branch(self):
 
601
        """Executed at the end of a try/finally to cleanup a bound branch.
 
602
 
 
603
        If the branch wasn't bound, this is a no-op.
 
604
        If it was, it resents self.branch to the local branch, instead
 
605
        of being the master.
 
606
        """
 
607
        if not self.bound_branch:
 
608
            return
 
609
        if self.master_locked:
 
610
            self.master_branch.unlock()
 
611
 
 
612
    def _escape_commit_message(self):
 
613
        """Replace xml-incompatible control characters."""
 
614
        # FIXME: RBC 20060419 this should be done by the revision
 
615
        # serialiser not by commit. Then we can also add an unescaper
 
616
        # in the deserializer and start roundtripping revision messages
 
617
        # precisely. See repository_implementations/test_repository.py
 
618
 
 
619
        # Python strings can include characters that can't be
 
620
        # represented in well-formed XML; escape characters that
 
621
        # aren't listed in the XML specification
 
622
        # (http://www.w3.org/TR/REC-xml/#NT-Char).
 
623
        self.message, escape_count = re.subn(
 
624
            u'[^\x09\x0A\x0D\u0020-\uD7FF\uE000-\uFFFD]+',
 
625
            lambda match: match.group(0).encode('unicode_escape'),
 
626
            self.message)
 
627
        if escape_count:
 
628
            self.reporter.escaped(escape_count, self.message)
 
629
 
643
630
    def _gather_parents(self):
644
631
        """Record the parents of a merge for merge detection."""
645
632
        # TODO: Make sure that this list doesn't contain duplicate
646
633
        # entries and the order is preserved when doing this.
647
 
        if self.use_record_iter_changes:
648
 
            return
649
 
        self.basis_inv = self.basis_tree.inventory
 
634
        self.parents = self.work_tree.get_parent_ids()
650
635
        self.parent_invs = [self.basis_inv]
651
636
        for revision in self.parents[1:]:
652
637
            if self.branch.repository.has_revision(revision):
659
644
    def _update_builder_with_changes(self):
660
645
        """Update the commit builder with the data about what has changed.
661
646
        """
 
647
        # Build the revision inventory.
 
648
        #
 
649
        # This starts by creating a new empty inventory. Depending on
 
650
        # which files are selected for commit, and what is present in the
 
651
        # current tree, the new inventory is populated. inventory entries
 
652
        # which are candidates for modification have their revision set to
 
653
        # None; inventory entries that are carried over untouched have their
 
654
        # revision set to their prior value.
 
655
        #
 
656
        # ESEPARATIONOFCONCERNS: this function is diffing and using the diff
 
657
        # results to create a new inventory at the same time, which results
 
658
        # in bugs like #46635.  Any reason not to use/enhance Tree.changes_from?
 
659
        # ADHB 11-07-2006
 
660
 
662
661
        exclude = self.exclude
663
 
        specific_files = self.specific_files
 
662
        specific_files = self.specific_files or []
664
663
        mutter("Selecting files for commit with filter %s", specific_files)
665
664
 
666
 
        self._check_strict()
667
 
        if self.use_record_iter_changes:
668
 
            iter_changes = self.work_tree.iter_changes(self.basis_tree,
669
 
                specific_files=specific_files)
670
 
            iter_changes = self._filter_iter_changes(iter_changes)
671
 
            for file_id, path, fs_hash in self.builder.record_iter_changes(
672
 
                self.work_tree, self.basis_revid, iter_changes):
673
 
                self.work_tree._observed_sha1(file_id, path, fs_hash)
674
 
        else:
675
 
            # Build the new inventory
676
 
            self._populate_from_inventory()
677
 
            self._record_unselected()
678
 
            self._report_and_accumulate_deletes()
679
 
 
680
 
    def _filter_iter_changes(self, iter_changes):
681
 
        """Process iter_changes.
682
 
 
683
 
        This method reports on the changes in iter_changes to the user, and 
684
 
        converts 'missing' entries in the iter_changes iterator to 'deleted'
685
 
        entries. 'missing' entries have their
686
 
 
687
 
        :param iter_changes: An iter_changes to process.
688
 
        :return: A generator of changes.
689
 
        """
690
 
        reporter = self.reporter
691
 
        report_changes = reporter.is_verbose()
692
 
        deleted_ids = []
693
 
        for change in iter_changes:
694
 
            if report_changes:
695
 
                old_path = change[1][0]
696
 
                new_path = change[1][1]
697
 
                versioned = change[3][1]
698
 
            kind = change[6][1]
699
 
            versioned = change[3][1]
700
 
            if kind is None and versioned:
701
 
                # 'missing' path
702
 
                if report_changes:
703
 
                    reporter.missing(new_path)
704
 
                deleted_ids.append(change[0])
705
 
                # Reset the new path (None) and new versioned flag (False)
706
 
                change = (change[0], (change[1][0], None), change[2],
707
 
                    (change[3][0], False)) + change[4:]
708
 
            elif kind == 'tree-reference':
709
 
                if self.recursive == 'down':
710
 
                    self._commit_nested_tree(change[0], change[1][1])
711
 
            if change[3][0] or change[3][1]:
712
 
                yield change
713
 
                if report_changes:
714
 
                    if new_path is None:
715
 
                        reporter.deleted(old_path)
716
 
                    elif old_path is None:
717
 
                        reporter.snapshot_change('added', new_path)
718
 
                    elif old_path != new_path:
719
 
                        reporter.renamed('renamed', old_path, new_path)
720
 
                    else:
721
 
                        if (new_path or 
722
 
                            self.work_tree.branch.repository._format.rich_root_data):
723
 
                            # Don't report on changes to '' in non rich root
724
 
                            # repositories.
725
 
                            reporter.snapshot_change('modified', new_path)
726
 
            self._next_progress_entry()
727
 
        # Unversion IDs that were found to be deleted
728
 
        self.deleted_ids = deleted_ids
729
 
 
730
 
    def _record_unselected(self):
 
665
        # Build the new inventory
 
666
        self._populate_from_inventory()
 
667
 
731
668
        # If specific files are selected, then all un-selected files must be
732
669
        # recorded in their previous state. For more details, see
733
670
        # https://lists.ubuntu.com/archives/bazaar/2007q3/028476.html.
734
 
        if self.specific_files or self.exclude:
735
 
            specific_files = self.specific_files or []
 
671
        if specific_files or exclude:
736
672
            for path, old_ie in self.basis_inv.iter_entries():
737
673
                if old_ie.file_id in self.builder.new_inventory:
738
674
                    # already added - skip.
739
675
                    continue
740
676
                if (is_inside_any(specific_files, path)
741
 
                    and not is_inside_any(self.exclude, path)):
 
677
                    and not is_inside_any(exclude, path)):
742
678
                    # was inside the selected path, and not excluded - if not
743
679
                    # present it has been deleted so skip.
744
680
                    continue
745
681
                # From here down it was either not selected, or was excluded:
 
682
                if old_ie.kind == 'directory':
 
683
                    self._next_progress_entry()
746
684
                # We preserve the entry unaltered.
747
685
                ie = old_ie.copy()
748
686
                # Note: specific file commits after a merge are currently
754
692
                    self.basis_tree, None)
755
693
 
756
694
    def _report_and_accumulate_deletes(self):
 
695
        # XXX: Could the list of deleted paths and ids be instead taken from
 
696
        # _populate_from_inventory?
757
697
        if (isinstance(self.basis_inv, Inventory)
758
698
            and isinstance(self.builder.new_inventory, Inventory)):
759
699
            # the older Inventory classes provide a _byid dict, and building a
777
717
                self.builder.record_delete(path, file_id)
778
718
                self.reporter.deleted(path)
779
719
 
780
 
    def _check_strict(self):
781
 
        # XXX: when we use iter_changes this would likely be faster if
782
 
        # iter_changes would check for us (even in the presence of
783
 
        # selected_files).
 
720
    def _populate_from_inventory(self):
 
721
        """Populate the CommitBuilder by walking the working tree inventory."""
784
722
        if self.strict:
785
723
            # raise an exception as soon as we find a single unknown.
786
724
            for unknown in self.work_tree.unknowns():
787
725
                raise StrictCommitFailed()
788
726
 
789
 
    def _populate_from_inventory(self):
790
 
        """Populate the CommitBuilder by walking the working tree inventory."""
791
 
        # Build the revision inventory.
792
 
        #
793
 
        # This starts by creating a new empty inventory. Depending on
794
 
        # which files are selected for commit, and what is present in the
795
 
        # current tree, the new inventory is populated. inventory entries
796
 
        # which are candidates for modification have their revision set to
797
 
        # None; inventory entries that are carried over untouched have their
798
 
        # revision set to their prior value.
799
 
        #
800
 
        # ESEPARATIONOFCONCERNS: this function is diffing and using the diff
801
 
        # results to create a new inventory at the same time, which results
802
 
        # in bugs like #46635.  Any reason not to use/enhance Tree.changes_from?
803
 
        # ADHB 11-07-2006
804
 
 
805
727
        specific_files = self.specific_files
806
728
        exclude = self.exclude
807
729
        report_changes = self.reporter.is_verbose()
821
743
            name = existing_ie.name
822
744
            parent_id = existing_ie.parent_id
823
745
            kind = existing_ie.kind
 
746
            if kind == 'directory':
 
747
                self._next_progress_entry()
824
748
            # Skip files that have been deleted from the working tree.
825
749
            # The deleted path ids are also recorded so they can be explicitly
826
750
            # unversioned later.
844
768
                # _update_builder_with_changes.
845
769
                continue
846
770
            content_summary = self.work_tree.path_content_summary(path)
847
 
            kind = content_summary[0]
848
771
            # Note that when a filter of specific files is given, we must only
849
772
            # skip/record deleted files matching that filter.
850
773
            if not specific_files or is_inside_any(specific_files, path):
851
 
                if kind == 'missing':
 
774
                if content_summary[0] == 'missing':
852
775
                    if not deleted_paths:
853
776
                        # path won't have been split yet.
854
777
                        path_segments = splitpath(path)
856
779
                    for segment in path_segments:
857
780
                        deleted_dict = deleted_dict.setdefault(segment, {})
858
781
                    self.reporter.missing(path)
859
 
                    self._next_progress_entry()
860
782
                    deleted_ids.append(file_id)
861
783
                    continue
862
784
            # TODO: have the builder do the nested commit just-in-time IF and
863
785
            # only if needed.
864
 
            if kind == 'tree-reference':
 
786
            if content_summary[0] == 'tree-reference':
865
787
                # enforce repository nested tree policy.
866
788
                if (not self.work_tree.supports_tree_reference() or
867
789
                    # repository does not support it either.
868
790
                    not self.branch.repository._format.supports_tree_reference):
869
 
                    kind = 'directory'
870
 
                    content_summary = (kind, None, None, None)
871
 
                elif self.recursive == 'down':
 
791
                    content_summary = ('directory',) + content_summary[1:]
 
792
            kind = content_summary[0]
 
793
            # TODO: specific_files filtering before nested tree processing
 
794
            if kind == 'tree-reference':
 
795
                if self.recursive == 'down':
872
796
                    nested_revision_id = self._commit_nested_tree(
873
797
                        file_id, path)
874
 
                    content_summary = (kind, None, None, nested_revision_id)
 
798
                    content_summary = content_summary[:3] + (
 
799
                        nested_revision_id,)
875
800
                else:
876
 
                    nested_revision_id = self.work_tree.get_reference_revision(file_id)
877
 
                    content_summary = (kind, None, None, nested_revision_id)
 
801
                    content_summary = content_summary[:3] + (
 
802
                        self.work_tree.get_reference_revision(file_id),)
878
803
 
879
804
            # Record an entry for this item
880
805
            # Note: I don't particularly want to have the existing_ie
886
811
                content_summary)
887
812
 
888
813
        # Unversion IDs that were found to be deleted
889
 
        self.deleted_ids = deleted_ids
 
814
        self.work_tree.unversion(deleted_ids)
890
815
 
891
816
    def _commit_nested_tree(self, file_id, path):
892
817
        "Commit a nested tree."
949
874
            InventoryEntry.MODIFIED_AND_RENAMED):
950
875
            old_path = self.basis_inv.id2path(ie.file_id)
951
876
            self.reporter.renamed(change, old_path, path)
952
 
            self._next_progress_entry()
953
877
        else:
954
 
            if change == 'unchanged':
955
 
                return
956
878
            self.reporter.snapshot_change(change, path)
957
 
            self._next_progress_entry()
958
879
 
959
 
    def _set_progress_stage(self, name, counter=False):
 
880
    def _set_progress_stage(self, name, entries_title=None):
960
881
        """Set the progress stage and emit an update to the progress bar."""
961
882
        self.pb_stage_name = name
962
883
        self.pb_stage_count += 1
963
 
        if counter:
 
884
        self.pb_entries_title = entries_title
 
885
        if entries_title is not None:
964
886
            self.pb_entries_count = 0
965
 
        else:
966
 
            self.pb_entries_count = None
 
887
            self.pb_entries_total = '?'
967
888
        self._emit_progress()
968
889
 
969
890
    def _next_progress_entry(self):
972
893
        self._emit_progress()
973
894
 
974
895
    def _emit_progress(self):
975
 
        if self.pb_entries_count is not None:
976
 
            text = "%s [%d] - Stage" % (self.pb_stage_name,
977
 
                self.pb_entries_count)
 
896
        if self.pb_entries_title:
 
897
            if self.pb_entries_total == '?':
 
898
                text = "%s [%s %d] - Stage" % (self.pb_stage_name,
 
899
                    self.pb_entries_title, self.pb_entries_count)
 
900
            else:
 
901
                text = "%s [%s %d/%s] - Stage" % (self.pb_stage_name,
 
902
                    self.pb_entries_title, self.pb_entries_count,
 
903
                    str(self.pb_entries_total))
978
904
        else:
979
 
            text = "%s - Stage" % (self.pb_stage_name, )
 
905
            text = "%s - Stage" % (self.pb_stage_name)
980
906
        self.pb.update(text, self.pb_stage_count, self.pb_stage_total)
981
907
 
982
 
    def _set_specific_file_ids(self):
983
 
        """populate self.specific_file_ids if we will use it."""
984
 
        if not self.use_record_iter_changes:
985
 
            # If provided, ensure the specified files are versioned
986
 
            if self.specific_files is not None:
987
 
                # Note: This routine is being called because it raises
988
 
                # PathNotVersionedError as a side effect of finding the IDs. We
989
 
                # later use the ids we found as input to the working tree
990
 
                # inventory iterator, so we only consider those ids rather than
991
 
                # examining the whole tree again.
992
 
                # XXX: Dont we have filter_unversioned to do this more
993
 
                # cheaply?
994
 
                self.specific_file_ids = tree.find_ids_across_trees(
995
 
                    self.specific_files, [self.basis_tree, self.work_tree])
996
 
            else:
997
 
                self.specific_file_ids = None