← Back to branch summary

~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/tests/test_fetch.py

https://launchpad.net/bugs/524184
NEWS section template into a separate file

Show diffs side-by-side

added added

removed removed

Lines of Context:
bzrlib/tests/test_fetch.py
12
12
#
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
 
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
 
17
17
import os
18
18
import re
22
22
from bzrlib import (
23
23
    bzrdir,
24
24
    errors,
 
25
    osutils,
25
26
    merge,
26
27
    repository,
 
28
    versionedfile,
27
29
    )
28
30
from bzrlib.branch import Branch
29
31
from bzrlib.bzrdir import BzrDir
36
38
from bzrlib.workingtree import WorkingTree
37
39
 
38
40
# These tests are a bit old; please instead add new tests into
39
 
# interrepository_implementations/ so they'll run on all relevant
 
41
# per_interrepository/ so they'll run on all relevant
40
42
# combinations.
41
43
 
42
44
 
45
47
 
46
48
def fetch_steps(self, br_a, br_b, writable_a):
47
49
    """A foreign test method for testing fetch locally and remotely."""
48
 
     
 
50
 
49
51
    # TODO RBC 20060201 make this a repository test.
50
52
    repo_b = br_b.repository
51
53
    self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
52
54
    self.assertTrue(repo_b.has_revision(br_a.revision_history()[2]))
53
55
    self.assertEquals(len(br_b.revision_history()), 7)
54
 
    self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[2])[0], 0)
 
56
    br_b.fetch(br_a, br_a.revision_history()[2])
55
57
    # branch.fetch is not supposed to alter the revision history
56
58
    self.assertEquals(len(br_b.revision_history()), 7)
57
59
    self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
58
60
 
59
61
    # fetching the next revision up in sample data copies one revision
60
 
    self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[3])[0], 1)
 
62
    br_b.fetch(br_a, br_a.revision_history()[3])
61
63
    self.assertTrue(repo_b.has_revision(br_a.revision_history()[3]))
62
64
    self.assertFalse(has_revision(br_a, br_b.revision_history()[6]))
63
65
    self.assertTrue(br_a.repository.has_revision(br_b.revision_history()[5]))
65
67
    # When a non-branch ancestor is missing, it should be unlisted...
66
68
    # as its not reference from the inventory weave.
67
69
    br_b4 = self.make_branch('br_4')
68
 
    count, failures = br_b4.fetch(br_b)
69
 
    self.assertEqual(count, 7)
70
 
    self.assertEqual(failures, [])
 
70
    br_b4.fetch(br_b)
71
71
 
72
 
    self.assertEqual(writable_a.fetch(br_b)[0], 1)
 
72
    writable_a.fetch(br_b)
73
73
    self.assertTrue(has_revision(br_a, br_b.revision_history()[3]))
74
74
    self.assertTrue(has_revision(br_a, br_b.revision_history()[4]))
75
 
        
 
75
 
76
76
    br_b2 = self.make_branch('br_b2')
77
 
    self.assertEquals(br_b2.fetch(br_b)[0], 7)
 
77
    br_b2.fetch(br_b)
78
78
    self.assertTrue(has_revision(br_b2, br_b.revision_history()[4]))
79
79
    self.assertTrue(has_revision(br_b2, br_a.revision_history()[2]))
80
80
    self.assertFalse(has_revision(br_b2, br_a.revision_history()[3]))
81
81
 
82
82
    br_a2 = self.make_branch('br_a2')
83
 
    self.assertEquals(br_a2.fetch(br_a)[0], 9)
 
83
    br_a2.fetch(br_a)
84
84
    self.assertTrue(has_revision(br_a2, br_b.revision_history()[4]))
85
85
    self.assertTrue(has_revision(br_a2, br_a.revision_history()[3]))
86
86
    self.assertTrue(has_revision(br_a2, br_a.revision_history()[2]))
87
87
 
88
88
    br_a3 = self.make_branch('br_a3')
89
 
    # pulling a branch with no revisions grabs nothing, regardless of 
 
89
    # pulling a branch with no revisions grabs nothing, regardless of
90
90
    # whats in the inventory.
91
 
    self.assertEquals(br_a3.fetch(br_a2)[0], 0)
 
91
    br_a3.fetch(br_a2)
92
92
    for revno in range(4):
93
93
        self.assertFalse(
94
94
            br_a3.repository.has_revision(br_a.revision_history()[revno]))
95
 
    self.assertEqual(br_a3.fetch(br_a2, br_a.revision_history()[2])[0], 3)
 
95
    br_a3.fetch(br_a2, br_a.revision_history()[2])
96
96
    # pull the 3 revisions introduced by a@u-0-3
97
 
    fetched = br_a3.fetch(br_a2, br_a.revision_history()[3])[0]
98
 
    self.assertEquals(fetched, 3, "fetched %d instead of 3" % fetched)
99
 
    # InstallFailed should be raised if the branch is missing the revision
 
97
    br_a3.fetch(br_a2, br_a.revision_history()[3])
 
98
    # NoSuchRevision should be raised if the branch is missing the revision
100
99
    # that was requested.
101
 
    self.assertRaises(errors.InstallFailed, br_a3.fetch, br_a2, 'pizza')
 
100
    self.assertRaises(errors.NoSuchRevision, br_a3.fetch, br_a2, 'pizza')
102
101
 
103
102
    # TODO: Test trying to fetch from a branch that points to a revision not
104
103
    # actually present in its repository.  Not every branch format allows you
107
106
    # every branch supports that.  -- mbp 20070814
108
107
 
109
108
    #TODO: test that fetch correctly does reweaving when needed. RBC 20051008
110
 
    # Note that this means - updating the weave when ghosts are filled in to 
 
109
    # Note that this means - updating the weave when ghosts are filled in to
111
110
    # add the right parents.
112
111
 
113
112
 
120
119
 
121
120
    def test_fetch_self(self):
122
121
        wt = self.make_branch_and_tree('br')
123
 
        self.assertEqual(wt.branch.fetch(wt.branch), (0, []))
 
122
        wt.branch.fetch(wt.branch)
124
123
 
125
124
    def test_fetch_root_knit(self):
126
125
        """Ensure that knit2.fetch() updates the root knit
127
 
        
 
126
 
128
127
        This tests the case where the root has a new revision, but there are no
129
128
        corresponding filename, parent, contents or other changes.
130
129
        """
170
169
        knit3_tree = self.make_branch_and_tree('knit3',
171
170
            format='dirstate-with-subtree')
172
171
        knit3_tree.commit('blah')
173
 
        self.assertRaises(errors.IncompatibleRepositories,
174
 
                          knit_tree.branch.fetch, knit3_tree.branch)
 
172
        e = self.assertRaises(errors.IncompatibleRepositories,
 
173
                              knit_tree.branch.fetch, knit3_tree.branch)
 
174
        self.assertContainsRe(str(e),
 
175
            r"(?m).*/knit.*\nis not compatible with\n.*/knit3/.*\n"
 
176
            r"different rich-root support")
175
177
 
176
178
 
177
179
class TestMergeFetch(TestCaseWithTransport):
279
281
        wt.commit("changed file")
280
282
        target = BzrDir.create_branch_and_repo("target/")
281
283
        source = Branch.open(self.get_readonly_url("source/"))
282
 
        self.assertEqual(target.fetch(source), (2, []))
283
 
        # this is the path to the literal file. As format changes 
 
284
        target.fetch(source)
 
285
        # this is the path to the literal file. As format changes
284
286
        # occur it needs to be updated. FIXME: ask the store for the
285
287
        # path.
286
288
        self.log("web server logs are:")
287
289
        http_logs = self.get_readonly_server().logs
288
290
        self.log('\n'.join(http_logs))
289
 
        # unfortunately this log entry is branch format specific. We could 
290
 
        # factor out the 'what files does this format use' to a method on the 
 
291
        # unfortunately this log entry is branch format specific. We could
 
292
        # factor out the 'what files does this format use' to a method on the
291
293
        # repository, which would let us to this generically. RBC 20060419
292
294
        # RBC 20080408: Or perhaps we can assert that no files are fully read
293
295
        # twice?
294
296
        self.assertEqual(1, self._count_log_matches('/ce/id.kndx', http_logs))
295
297
        self.assertEqual(1, self._count_log_matches('/ce/id.knit', http_logs))
296
298
        self.assertEqual(1, self._count_log_matches('inventory.kndx', http_logs))
297
 
        # this r-h check test will prevent regressions, but it currently already 
 
299
        # this r-h check test will prevent regressions, but it currently already
298
300
        # passes, before the patch to cache-rh is applied :[
299
301
        self.assertTrue(1 >= self._count_log_matches('revision-history',
300
302
                                                     http_logs))
310
312
        source = Branch.open(
311
313
            self.get_readonly_url("source/"),
312
314
            possible_transports=[source.bzrdir.root_transport])
313
 
        self.assertEqual(target.fetch(source), (0, []))
 
315
        target.fetch(source)
314
316
        # should make just two requests
315
317
        http_logs = self.get_readonly_server().logs
316
318
        self.log("web server logs are:")
319
321
        self.assertEqual(1, self._count_log_matches('branch/format', http_logs))
320
322
        self.assertEqual(1, self._count_log_matches('repository/format',
321
323
            http_logs))
 
324
        self.assertEqual(1, self._count_log_matches('revisions.kndx',
 
325
            http_logs))
322
326
        self.assertTrue(1 >= self._count_log_matches('revision-history',
323
327
                                                     http_logs))
324
328
        self.assertTrue(1 >= self._count_log_matches('last-revision',
325
329
                                                     http_logs))
326
 
        self.assertEqual(4, len(http_logs))
 
330
        self.assertLength(5, http_logs)
 
331
 
 
332
 
 
333
class TestKnitToPackFetch(TestCaseWithTransport):
 
334
 
 
335
    def find_get_record_stream(self, calls, expected_count=1):
 
336
        """In a list of calls, find the last 'get_record_stream'.
 
337
 
 
338
        :param expected_count: The number of calls we should exepect to find.
 
339
            If a different number is found, an assertion is raised.
 
340
        """
 
341
        get_record_call = None
 
342
        call_count = 0
 
343
        for call in calls:
 
344
            if call[0] == 'get_record_stream':
 
345
                call_count += 1
 
346
                get_record_call = call
 
347
        self.assertEqual(expected_count, call_count)
 
348
        return get_record_call
 
349
 
 
350
    def test_fetch_with_deltas_no_delta_closure(self):
 
351
        tree = self.make_branch_and_tree('source', format='dirstate')
 
352
        target = self.make_repository('target', format='pack-0.92')
 
353
        self.build_tree(['source/file'])
 
354
        tree.set_root_id('root-id')
 
355
        tree.add('file', 'file-id')
 
356
        tree.commit('one', rev_id='rev-one')
 
357
        source = tree.branch.repository
 
358
        source.texts = versionedfile.RecordingVersionedFilesDecorator(
 
359
                        source.texts)
 
360
        source.signatures = versionedfile.RecordingVersionedFilesDecorator(
 
361
                        source.signatures)
 
362
        source.revisions = versionedfile.RecordingVersionedFilesDecorator(
 
363
                        source.revisions)
 
364
        source.inventories = versionedfile.RecordingVersionedFilesDecorator(
 
365
                        source.inventories)
 
366
        # precondition
 
367
        self.assertTrue(target._format._fetch_uses_deltas)
 
368
        target.fetch(source, revision_id='rev-one')
 
369
        self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
 
370
                          target._format._fetch_order, False),
 
371
                         self.find_get_record_stream(source.texts.calls))
 
372
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
373
          target._format._fetch_order, False),
 
374
          self.find_get_record_stream(source.inventories.calls, 2))
 
375
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
376
                          target._format._fetch_order, False),
 
377
                         self.find_get_record_stream(source.revisions.calls))
 
378
        # XXX: Signatures is special, and slightly broken. The
 
379
        # standard item_keys_introduced_by actually does a lookup for every
 
380
        # signature to see if it exists, rather than waiting to do them all at
 
381
        # once at the end. The fetch code then does an all-at-once and just
 
382
        # allows for some of them to be missing.
 
383
        # So we know there will be extra calls, but the *last* one is the one
 
384
        # we care about.
 
385
        signature_calls = source.signatures.calls[-1:]
 
386
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
387
                          target._format._fetch_order, False),
 
388
                         self.find_get_record_stream(signature_calls))
 
389
 
 
390
    def test_fetch_no_deltas_with_delta_closure(self):
 
391
        tree = self.make_branch_and_tree('source', format='dirstate')
 
392
        target = self.make_repository('target', format='pack-0.92')
 
393
        self.build_tree(['source/file'])
 
394
        tree.set_root_id('root-id')
 
395
        tree.add('file', 'file-id')
 
396
        tree.commit('one', rev_id='rev-one')
 
397
        source = tree.branch.repository
 
398
        source.texts = versionedfile.RecordingVersionedFilesDecorator(
 
399
                        source.texts)
 
400
        source.signatures = versionedfile.RecordingVersionedFilesDecorator(
 
401
                        source.signatures)
 
402
        source.revisions = versionedfile.RecordingVersionedFilesDecorator(
 
403
                        source.revisions)
 
404
        source.inventories = versionedfile.RecordingVersionedFilesDecorator(
 
405
                        source.inventories)
 
406
        # XXX: This won't work in general, but for the dirstate format it does.
 
407
        old_fetch_uses_deltas_setting = target._format._fetch_uses_deltas
 
408
        def restore():
 
409
            target._format._fetch_uses_deltas = old_fetch_uses_deltas_setting
 
410
        self.addCleanup(restore)
 
411
        target._format._fetch_uses_deltas = False
 
412
        target.fetch(source, revision_id='rev-one')
 
413
        self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
 
414
                          target._format._fetch_order, True),
 
415
                         self.find_get_record_stream(source.texts.calls))
 
416
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
417
            target._format._fetch_order, True),
 
418
            self.find_get_record_stream(source.inventories.calls, 2))
 
419
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
420
                          target._format._fetch_order, True),
 
421
                         self.find_get_record_stream(source.revisions.calls))
 
422
        # XXX: Signatures is special, and slightly broken. The
 
423
        # standard item_keys_introduced_by actually does a lookup for every
 
424
        # signature to see if it exists, rather than waiting to do them all at
 
425
        # once at the end. The fetch code then does an all-at-once and just
 
426
        # allows for some of them to be missing.
 
427
        # So we know there will be extra calls, but the *last* one is the one
 
428
        # we care about.
 
429
        signature_calls = source.signatures.calls[-1:]
 
430
        self.assertEqual(('get_record_stream', [('rev-one',)],
 
431
                          target._format._fetch_order, True),
 
432
                         self.find_get_record_stream(signature_calls))
 
433
 
 
434
    def test_fetch_revisions_with_deltas_into_pack(self):
 
435
        # See BUG #261339, dev versions of bzr could accidentally create deltas
 
436
        # in revision texts in knit branches (when fetching from packs). So we
 
437
        # ensure that *if* a knit repository has a delta in revisions, that it
 
438
        # gets properly expanded back into a fulltext when stored in the pack
 
439
        # file.
 
440
        tree = self.make_branch_and_tree('source', format='dirstate')
 
441
        target = self.make_repository('target', format='pack-0.92')
 
442
        self.build_tree(['source/file'])
 
443
        tree.set_root_id('root-id')
 
444
        tree.add('file', 'file-id')
 
445
        tree.commit('one', rev_id='rev-one')
 
446
        # Hack the KVF for revisions so that it "accidentally" allows a delta
 
447
        tree.branch.repository.revisions._max_delta_chain = 200
 
448
        tree.commit('two', rev_id='rev-two')
 
449
        source = tree.branch.repository
 
450
        # Ensure that we stored a delta
 
451
        source.lock_read()
 
452
        self.addCleanup(source.unlock)
 
453
        record = source.revisions.get_record_stream([('rev-two',)],
 
454
            'unordered', False).next()
 
455
        self.assertEqual('knit-delta-gz', record.storage_kind)
 
456
        target.fetch(tree.branch.repository, revision_id='rev-two')
 
457
        # The record should get expanded back to a fulltext
 
458
        target.lock_read()
 
459
        self.addCleanup(target.unlock)
 
460
        record = target.revisions.get_record_stream([('rev-two',)],
 
461
            'unordered', False).next()
 
462
        self.assertEqual('knit-ft-gz', record.storage_kind)
 
463
 
 
464
    def test_fetch_with_fallback_and_merge(self):
 
465
        builder = self.make_branch_builder('source', format='pack-0.92')
 
466
        builder.start_series()
 
467
        # graph
 
468
        #   A
 
469
        #   |\
 
470
        #   B C
 
471
        #   | |
 
472
        #   | D
 
473
        #   | |
 
474
        #   | E
 
475
        #    \|
 
476
        #     F
 
477
        # A & B are present in the base (stacked-on) repository, A-E are
 
478
        # present in the source.
 
479
        # This reproduces bug #304841
 
480
        # We need a large enough inventory that total size of compressed deltas
 
481
        # is shorter than the size of a compressed fulltext. We have to use
 
482
        # random ids because otherwise the inventory fulltext compresses too
 
483
        # well and the deltas get bigger.
 
484
        to_add = [
 
485
            ('add', ('', 'TREE_ROOT', 'directory', None))]
 
486
        for i in xrange(10):
 
487
            fname = 'file%03d' % (i,)
 
488
            fileid = '%s-%s' % (fname, osutils.rand_chars(64))
 
489
            to_add.append(('add', (fname, fileid, 'file', 'content\n')))
 
490
        builder.build_snapshot('A', None, to_add)
 
491
        builder.build_snapshot('B', ['A'], [])
 
492
        builder.build_snapshot('C', ['A'], [])
 
493
        builder.build_snapshot('D', ['C'], [])
 
494
        builder.build_snapshot('E', ['D'], [])
 
495
        builder.build_snapshot('F', ['E', 'B'], [])
 
496
        builder.finish_series()
 
497
        source_branch = builder.get_branch()
 
498
        source_branch.bzrdir.sprout('base', revision_id='B')
 
499
        target_branch = self.make_branch('target', format='1.6')
 
500
        target_branch.set_stacked_on_url('../base')
 
501
        source = source_branch.repository
 
502
        source.lock_read()
 
503
        self.addCleanup(source.unlock)
 
504
        source.inventories = versionedfile.OrderingVersionedFilesDecorator(
 
505
                        source.inventories,
 
506
                        key_priority={('E',): 1, ('D',): 2, ('C',): 4,
 
507
                                      ('F',): 3})
 
508
        # Ensure that the content is yielded in the proper order, and given as
 
509
        # the expected kinds
 
510
        records = [(record.key, record.storage_kind)
 
511
                   for record in source.inventories.get_record_stream(
 
512
                        [('D',), ('C',), ('E',), ('F',)], 'unordered', False)]
 
513
        self.assertEqual([(('E',), 'knit-delta-gz'), (('D',), 'knit-delta-gz'),
 
514
                          (('F',), 'knit-delta-gz'), (('C',), 'knit-delta-gz')],
 
515
                          records)
 
516
 
 
517
        target_branch.lock_write()
 
518
        self.addCleanup(target_branch.unlock)
 
519
        target = target_branch.repository
 
520
        target.fetch(source, revision_id='F')
 
521
        # 'C' should be expanded to a fulltext, but D and E should still be
 
522
        # deltas
 
523
        stream = target.inventories.get_record_stream(
 
524
            [('C',), ('D',), ('E',), ('F',)],
 
525
            'unordered', False)
 
526
        kinds = dict((record.key, record.storage_kind) for record in stream)
 
527
        self.assertEqual({('C',): 'knit-ft-gz', ('D',): 'knit-delta-gz',
 
528
                          ('E',): 'knit-delta-gz', ('F',): 'knit-delta-gz'},
 
529
                         kinds)
327
530
 
328
531
 
329
532
class Test1To2Fetch(TestCaseWithTransport):
375
578
        self.repo.fetch(self.tree.branch.repository, 'second-id')
376
579
        root_id = self.tree.get_root_id()
377
580
        self.assertEqual(
378
 
            ((root_id, 'left-parent'), (root_id, 'ghost-parent'),
379
 
             (root_id, 'not-ghost-parent')),
 
581
            ((root_id, 'left-parent'), (root_id, 'not-ghost-parent')),
380
582
            self.get_parents(root_id, 'second-id'))
381
583
 
382
584
    def make_two_commits(self, change_root, fetch_twice):