~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/tests/test_fetch.py

  • Committer: John Arbash Meinel
  • Author(s): Mark Hammond
  • Date: 2008-09-09 17:02:21 UTC
  • mto: This revision was merged to the branch mainline in revision 3697.
  • Revision ID: john@arbash-meinel.com-20080909170221-svim3jw2mrz0amp3
An updated transparent icon for bzr.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2005-2011, 2016 Canonical Ltd
 
1
# Copyright (C) 2005, 2007 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
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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
 
 
 
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 
16
 
 
17
import os
 
18
import re
 
19
import sys
 
20
 
 
21
import bzrlib
17
22
from bzrlib import (
18
23
    bzrdir,
19
24
    errors,
20
 
    osutils,
21
 
    revision as _mod_revision,
 
25
    merge,
 
26
    repository,
22
27
    versionedfile,
23
28
    )
24
29
from bzrlib.branch import Branch
 
30
from bzrlib.bzrdir import BzrDir
25
31
from bzrlib.repofmt import knitrepo
26
32
from bzrlib.tests import TestCaseWithTransport
 
33
from bzrlib.tests.http_utils import TestCaseWithWebserver
27
34
from bzrlib.tests.test_revision import make_branches
 
35
from bzrlib.trace import mutter
28
36
from bzrlib.upgrade import Convert
29
37
from bzrlib.workingtree import WorkingTree
30
38
 
31
39
# These tests are a bit old; please instead add new tests into
32
 
# per_interrepository/ so they'll run on all relevant
 
40
# interrepository_implementations/ so they'll run on all relevant
33
41
# combinations.
34
42
 
35
43
 
36
44
def has_revision(branch, revision_id):
37
45
    return branch.repository.has_revision(revision_id)
38
46
 
39
 
 
40
 
def revision_history(branch):
41
 
    branch.lock_read()
42
 
    try:
43
 
        graph = branch.repository.get_graph()
44
 
        history = list(graph.iter_lefthand_ancestry(branch.last_revision(),
45
 
            [_mod_revision.NULL_REVISION]))
46
 
    finally:
47
 
        branch.unlock()
48
 
    history.reverse()
49
 
    return history
50
 
 
51
 
 
52
47
def fetch_steps(self, br_a, br_b, writable_a):
53
48
    """A foreign test method for testing fetch locally and remotely."""
54
 
 
 
49
     
55
50
    # TODO RBC 20060201 make this a repository test.
56
51
    repo_b = br_b.repository
57
 
    self.assertFalse(repo_b.has_revision(revision_history(br_a)[3]))
58
 
    self.assertTrue(repo_b.has_revision(revision_history(br_a)[2]))
59
 
    self.assertEqual(len(revision_history(br_b)), 7)
60
 
    br_b.fetch(br_a, revision_history(br_a)[2])
 
52
    self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
 
53
    self.assertTrue(repo_b.has_revision(br_a.revision_history()[2]))
 
54
    self.assertEquals(len(br_b.revision_history()), 7)
 
55
    self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[2])[0], 0)
61
56
    # branch.fetch is not supposed to alter the revision history
62
 
    self.assertEqual(len(revision_history(br_b)), 7)
63
 
    self.assertFalse(repo_b.has_revision(revision_history(br_a)[3]))
 
57
    self.assertEquals(len(br_b.revision_history()), 7)
 
58
    self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
64
59
 
65
60
    # fetching the next revision up in sample data copies one revision
66
 
    br_b.fetch(br_a, revision_history(br_a)[3])
67
 
    self.assertTrue(repo_b.has_revision(revision_history(br_a)[3]))
68
 
    self.assertFalse(has_revision(br_a, revision_history(br_b)[6]))
69
 
    self.assertTrue(br_a.repository.has_revision(revision_history(br_b)[5]))
 
61
    self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[3])[0], 1)
 
62
    self.assertTrue(repo_b.has_revision(br_a.revision_history()[3]))
 
63
    self.assertFalse(has_revision(br_a, br_b.revision_history()[6]))
 
64
    self.assertTrue(br_a.repository.has_revision(br_b.revision_history()[5]))
70
65
 
71
66
    # When a non-branch ancestor is missing, it should be unlisted...
72
67
    # as its not reference from the inventory weave.
73
68
    br_b4 = self.make_branch('br_4')
74
 
    br_b4.fetch(br_b)
75
 
 
76
 
    writable_a.fetch(br_b)
77
 
    self.assertTrue(has_revision(br_a, revision_history(br_b)[3]))
78
 
    self.assertTrue(has_revision(br_a, revision_history(br_b)[4]))
79
 
 
 
69
    count, failures = br_b4.fetch(br_b)
 
70
    self.assertEqual(count, 7)
 
71
    self.assertEqual(failures, [])
 
72
 
 
73
    self.assertEqual(writable_a.fetch(br_b)[0], 1)
 
74
    self.assertTrue(has_revision(br_a, br_b.revision_history()[3]))
 
75
    self.assertTrue(has_revision(br_a, br_b.revision_history()[4]))
 
76
        
80
77
    br_b2 = self.make_branch('br_b2')
81
 
    br_b2.fetch(br_b)
82
 
    self.assertTrue(has_revision(br_b2, revision_history(br_b)[4]))
83
 
    self.assertTrue(has_revision(br_b2, revision_history(br_a)[2]))
84
 
    self.assertFalse(has_revision(br_b2, revision_history(br_a)[3]))
 
78
    self.assertEquals(br_b2.fetch(br_b)[0], 7)
 
79
    self.assertTrue(has_revision(br_b2, br_b.revision_history()[4]))
 
80
    self.assertTrue(has_revision(br_b2, br_a.revision_history()[2]))
 
81
    self.assertFalse(has_revision(br_b2, br_a.revision_history()[3]))
85
82
 
86
83
    br_a2 = self.make_branch('br_a2')
87
 
    br_a2.fetch(br_a)
88
 
    self.assertTrue(has_revision(br_a2, revision_history(br_b)[4]))
89
 
    self.assertTrue(has_revision(br_a2, revision_history(br_a)[3]))
90
 
    self.assertTrue(has_revision(br_a2, revision_history(br_a)[2]))
 
84
    self.assertEquals(br_a2.fetch(br_a)[0], 9)
 
85
    self.assertTrue(has_revision(br_a2, br_b.revision_history()[4]))
 
86
    self.assertTrue(has_revision(br_a2, br_a.revision_history()[3]))
 
87
    self.assertTrue(has_revision(br_a2, br_a.revision_history()[2]))
91
88
 
92
89
    br_a3 = self.make_branch('br_a3')
93
 
    # pulling a branch with no revisions grabs nothing, regardless of
 
90
    # pulling a branch with no revisions grabs nothing, regardless of 
94
91
    # whats in the inventory.
95
 
    br_a3.fetch(br_a2)
 
92
    self.assertEquals(br_a3.fetch(br_a2)[0], 0)
96
93
    for revno in range(4):
97
94
        self.assertFalse(
98
 
            br_a3.repository.has_revision(revision_history(br_a)[revno]))
99
 
    br_a3.fetch(br_a2, revision_history(br_a)[2])
 
95
            br_a3.repository.has_revision(br_a.revision_history()[revno]))
 
96
    self.assertEqual(br_a3.fetch(br_a2, br_a.revision_history()[2])[0], 3)
100
97
    # pull the 3 revisions introduced by a@u-0-3
101
 
    br_a3.fetch(br_a2, revision_history(br_a)[3])
102
 
    # NoSuchRevision should be raised if the branch is missing the revision
 
98
    fetched = br_a3.fetch(br_a2, br_a.revision_history()[3])[0]
 
99
    self.assertEquals(fetched, 3, "fetched %d instead of 3" % fetched)
 
100
    # InstallFailed should be raised if the branch is missing the revision
103
101
    # that was requested.
104
 
    self.assertRaises(errors.NoSuchRevision, br_a3.fetch, br_a2, 'pizza')
 
102
    self.assertRaises(errors.InstallFailed, br_a3.fetch, br_a2, 'pizza')
105
103
 
106
104
    # TODO: Test trying to fetch from a branch that points to a revision not
107
105
    # actually present in its repository.  Not every branch format allows you
110
108
    # every branch supports that.  -- mbp 20070814
111
109
 
112
110
    #TODO: test that fetch correctly does reweaving when needed. RBC 20051008
113
 
    # Note that this means - updating the weave when ghosts are filled in to
 
111
    # Note that this means - updating the weave when ghosts are filled in to 
114
112
    # add the right parents.
115
113
 
116
114
 
123
121
 
124
122
    def test_fetch_self(self):
125
123
        wt = self.make_branch_and_tree('br')
126
 
        wt.branch.fetch(wt.branch)
 
124
        self.assertEqual(wt.branch.fetch(wt.branch), (0, []))
127
125
 
128
126
    def test_fetch_root_knit(self):
129
127
        """Ensure that knit2.fetch() updates the root knit
130
 
 
 
128
        
131
129
        This tests the case where the root has a new revision, but there are no
132
130
        corresponding filename, parent, contents or other changes.
133
131
        """
252
250
                    rev_id).get_file_text('this-file-id'), text)
253
251
 
254
252
 
 
253
class TestHttpFetch(TestCaseWithWebserver):
 
254
    # FIXME RBC 20060124 this really isn't web specific, perhaps an
 
255
    # instrumented readonly transport? Can we do an instrumented
 
256
    # adapter and use self.get_readonly_url ?
 
257
 
 
258
    def test_fetch(self):
 
259
        #highest indices a: 5, b: 7
 
260
        br_a, br_b = make_branches(self)
 
261
        br_rem_a = Branch.open(self.get_readonly_url('branch1'))
 
262
        fetch_steps(self, br_rem_a, br_b, br_a)
 
263
 
 
264
    def _count_log_matches(self, target, logs):
 
265
        """Count the number of times the target file pattern was fetched in an http log"""
 
266
        get_succeeds_re = re.compile(
 
267
            '.*"GET .*%s HTTP/1.1" 20[06] - "-" "bzr/%s' %
 
268
            (     target,                    bzrlib.__version__))
 
269
        c = 0
 
270
        for line in logs:
 
271
            if get_succeeds_re.match(line):
 
272
                c += 1
 
273
        return c
 
274
 
 
275
    def test_weaves_are_retrieved_once(self):
 
276
        self.build_tree(("source/", "source/file", "target/"))
 
277
        # This test depends on knit dasta storage.
 
278
        wt = self.make_branch_and_tree('source', format='dirstate-tags')
 
279
        branch = wt.branch
 
280
        wt.add(["file"], ["id"])
 
281
        wt.commit("added file")
 
282
        open("source/file", 'w').write("blah\n")
 
283
        wt.commit("changed file")
 
284
        target = BzrDir.create_branch_and_repo("target/")
 
285
        source = Branch.open(self.get_readonly_url("source/"))
 
286
        self.assertEqual(target.fetch(source), (2, []))
 
287
        # this is the path to the literal file. As format changes 
 
288
        # occur it needs to be updated. FIXME: ask the store for the
 
289
        # path.
 
290
        self.log("web server logs are:")
 
291
        http_logs = self.get_readonly_server().logs
 
292
        self.log('\n'.join(http_logs))
 
293
        # unfortunately this log entry is branch format specific. We could 
 
294
        # factor out the 'what files does this format use' to a method on the 
 
295
        # repository, which would let us to this generically. RBC 20060419
 
296
        # RBC 20080408: Or perhaps we can assert that no files are fully read
 
297
        # twice?
 
298
        self.assertEqual(1, self._count_log_matches('/ce/id.kndx', http_logs))
 
299
        self.assertEqual(1, self._count_log_matches('/ce/id.knit', http_logs))
 
300
        self.assertEqual(1, self._count_log_matches('inventory.kndx', http_logs))
 
301
        # this r-h check test will prevent regressions, but it currently already 
 
302
        # passes, before the patch to cache-rh is applied :[
 
303
        self.assertTrue(1 >= self._count_log_matches('revision-history',
 
304
                                                     http_logs))
 
305
        self.assertTrue(1 >= self._count_log_matches('last-revision',
 
306
                                                     http_logs))
 
307
        # FIXME naughty poking in there.
 
308
        self.get_readonly_server().logs = []
 
309
        # check there is nothing more to fetch.  We take care to re-use the
 
310
        # existing transport so that the request logs we're about to examine
 
311
        # aren't cluttered with redundant probes for a smart server.
 
312
        # XXX: Perhaps this further parameterisation: test http with smart
 
313
        # server, and test http without smart server?
 
314
        source = Branch.open(
 
315
            self.get_readonly_url("source/"),
 
316
            possible_transports=[source.bzrdir.root_transport])
 
317
        self.assertEqual(target.fetch(source), (0, []))
 
318
        # should make just two requests
 
319
        http_logs = self.get_readonly_server().logs
 
320
        self.log("web server logs are:")
 
321
        self.log('\n'.join(http_logs))
 
322
        self.assertEqual(1, self._count_log_matches('branch-format', http_logs))
 
323
        self.assertEqual(1, self._count_log_matches('branch/format', http_logs))
 
324
        self.assertEqual(1, self._count_log_matches('repository/format',
 
325
            http_logs))
 
326
        self.assertTrue(1 >= self._count_log_matches('revision-history',
 
327
                                                     http_logs))
 
328
        self.assertTrue(1 >= self._count_log_matches('last-revision',
 
329
                                                     http_logs))
 
330
        self.assertEqual(4, len(http_logs))
 
331
 
 
332
 
255
333
class TestKnitToPackFetch(TestCaseWithTransport):
256
334
 
257
 
    def find_get_record_stream(self, calls, expected_count=1):
258
 
        """In a list of calls, find the last 'get_record_stream'.
 
335
    def find_get_record_stream(self, calls):
 
336
        """In a list of calls, find 'get_record_stream' calls.
259
337
 
260
 
        :param expected_count: The number of calls we should exepect to find.
261
 
            If a different number is found, an assertion is raised.
 
338
        This also ensures that there is only one get_record_stream call.
262
339
        """
263
340
        get_record_call = None
264
 
        call_count = 0
265
341
        for call in calls:
266
342
            if call[0] == 'get_record_stream':
267
 
                call_count += 1
 
343
                self.assertIs(None, get_record_call,
 
344
                              "there should only be one call to"
 
345
                              " get_record_stream")
268
346
                get_record_call = call
269
 
        self.assertEqual(expected_count, call_count)
 
347
        self.assertIsNot(None, get_record_call,
 
348
                         "there should be exactly one call to "
 
349
                         " get_record_stream")
270
350
        return get_record_call
271
351
 
272
352
    def test_fetch_with_deltas_no_delta_closure(self):
286
366
        source.inventories = versionedfile.RecordingVersionedFilesDecorator(
287
367
                        source.inventories)
288
368
        # precondition
289
 
        self.assertTrue(target._format._fetch_uses_deltas)
 
369
        self.assertTrue(target._fetch_uses_deltas)
290
370
        target.fetch(source, revision_id='rev-one')
291
371
        self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
292
 
                          target._format._fetch_order, False),
 
372
                          target._fetch_order, False),
293
373
                         self.find_get_record_stream(source.texts.calls))
294
374
        self.assertEqual(('get_record_stream', [('rev-one',)],
295
 
          target._format._fetch_order, False),
296
 
          self.find_get_record_stream(source.inventories.calls, 2))
 
375
                          target._fetch_order, False),
 
376
                         self.find_get_record_stream(source.inventories.calls))
 
377
        # Because of bugs in the old fetch code, revisions could accidentally
 
378
        # have deltas present in knits. However, it was never intended, so we
 
379
        # always for include_delta_closure=True, to make sure we get fulltexts.
 
380
        # bug #261339
297
381
        self.assertEqual(('get_record_stream', [('rev-one',)],
298
 
                          target._format._fetch_order, False),
 
382
                          target._fetch_order, True),
299
383
                         self.find_get_record_stream(source.revisions.calls))
300
384
        # XXX: Signatures is special, and slightly broken. The
301
385
        # standard item_keys_introduced_by actually does a lookup for every
306
390
        # we care about.
307
391
        signature_calls = source.signatures.calls[-1:]
308
392
        self.assertEqual(('get_record_stream', [('rev-one',)],
309
 
                          target._format._fetch_order, False),
 
393
                          target._fetch_order, True),
310
394
                         self.find_get_record_stream(signature_calls))
311
395
 
312
396
    def test_fetch_no_deltas_with_delta_closure(self):
325
409
                        source.revisions)
326
410
        source.inventories = versionedfile.RecordingVersionedFilesDecorator(
327
411
                        source.inventories)
328
 
        # XXX: This won't work in general, but for the dirstate format it does.
329
 
        self.overrideAttr(target._format, '_fetch_uses_deltas', False)
 
412
        target._fetch_uses_deltas = False
330
413
        target.fetch(source, revision_id='rev-one')
331
414
        self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
332
 
                          target._format._fetch_order, True),
 
415
                          target._fetch_order, True),
333
416
                         self.find_get_record_stream(source.texts.calls))
334
417
        self.assertEqual(('get_record_stream', [('rev-one',)],
335
 
            target._format._fetch_order, True),
336
 
            self.find_get_record_stream(source.inventories.calls, 2))
 
418
                          target._fetch_order, True),
 
419
                         self.find_get_record_stream(source.inventories.calls))
337
420
        self.assertEqual(('get_record_stream', [('rev-one',)],
338
 
                          target._format._fetch_order, True),
 
421
                          target._fetch_order, True),
339
422
                         self.find_get_record_stream(source.revisions.calls))
340
423
        # XXX: Signatures is special, and slightly broken. The
341
424
        # standard item_keys_introduced_by actually does a lookup for every
346
429
        # we care about.
347
430
        signature_calls = source.signatures.calls[-1:]
348
431
        self.assertEqual(('get_record_stream', [('rev-one',)],
349
 
                          target._format._fetch_order, True),
 
432
                          target._fetch_order, True),
350
433
                         self.find_get_record_stream(signature_calls))
351
434
 
352
 
    def test_fetch_revisions_with_deltas_into_pack(self):
353
 
        # See BUG #261339, dev versions of bzr could accidentally create deltas
354
 
        # in revision texts in knit branches (when fetching from packs). So we
355
 
        # ensure that *if* a knit repository has a delta in revisions, that it
356
 
        # gets properly expanded back into a fulltext when stored in the pack
357
 
        # file.
358
 
        tree = self.make_branch_and_tree('source', format='dirstate')
359
 
        target = self.make_repository('target', format='pack-0.92')
360
 
        self.build_tree(['source/file'])
361
 
        tree.set_root_id('root-id')
362
 
        tree.add('file', 'file-id')
363
 
        tree.commit('one', rev_id='rev-one')
364
 
        # Hack the KVF for revisions so that it "accidentally" allows a delta
365
 
        tree.branch.repository.revisions._max_delta_chain = 200
366
 
        tree.commit('two', rev_id='rev-two')
367
 
        source = tree.branch.repository
368
 
        # Ensure that we stored a delta
369
 
        source.lock_read()
370
 
        self.addCleanup(source.unlock)
371
 
        record = source.revisions.get_record_stream([('rev-two',)],
372
 
            'unordered', False).next()
373
 
        self.assertEqual('knit-delta-gz', record.storage_kind)
374
 
        target.fetch(tree.branch.repository, revision_id='rev-two')
375
 
        # The record should get expanded back to a fulltext
376
 
        target.lock_read()
377
 
        self.addCleanup(target.unlock)
378
 
        record = target.revisions.get_record_stream([('rev-two',)],
379
 
            'unordered', False).next()
380
 
        self.assertEqual('knit-ft-gz', record.storage_kind)
381
 
 
382
 
    def test_fetch_with_fallback_and_merge(self):
383
 
        builder = self.make_branch_builder('source', format='pack-0.92')
384
 
        builder.start_series()
385
 
        # graph
386
 
        #   A
387
 
        #   |\
388
 
        #   B C
389
 
        #   | |
390
 
        #   | D
391
 
        #   | |
392
 
        #   | E
393
 
        #    \|
394
 
        #     F
395
 
        # A & B are present in the base (stacked-on) repository, A-E are
396
 
        # present in the source.
397
 
        # This reproduces bug #304841
398
 
        # We need a large enough inventory that total size of compressed deltas
399
 
        # is shorter than the size of a compressed fulltext. We have to use
400
 
        # random ids because otherwise the inventory fulltext compresses too
401
 
        # well and the deltas get bigger.
402
 
        to_add = [
403
 
            ('add', ('', 'TREE_ROOT', 'directory', None))]
404
 
        for i in xrange(10):
405
 
            fname = 'file%03d' % (i,)
406
 
            fileid = '%s-%s' % (fname, osutils.rand_chars(64))
407
 
            to_add.append(('add', (fname, fileid, 'file', 'content\n')))
408
 
        builder.build_snapshot('A', None, to_add)
409
 
        builder.build_snapshot('B', ['A'], [])
410
 
        builder.build_snapshot('C', ['A'], [])
411
 
        builder.build_snapshot('D', ['C'], [])
412
 
        builder.build_snapshot('E', ['D'], [])
413
 
        builder.build_snapshot('F', ['E', 'B'], [])
414
 
        builder.finish_series()
415
 
        source_branch = builder.get_branch()
416
 
        source_branch.bzrdir.sprout('base', revision_id='B')
417
 
        target_branch = self.make_branch('target', format='1.6')
418
 
        target_branch.set_stacked_on_url('../base')
419
 
        source = source_branch.repository
420
 
        source.lock_read()
421
 
        self.addCleanup(source.unlock)
422
 
        source.inventories = versionedfile.OrderingVersionedFilesDecorator(
423
 
                        source.inventories,
424
 
                        key_priority={('E',): 1, ('D',): 2, ('C',): 4,
425
 
                                      ('F',): 3})
426
 
        # Ensure that the content is yielded in the proper order, and given as
427
 
        # the expected kinds
428
 
        records = [(record.key, record.storage_kind)
429
 
                   for record in source.inventories.get_record_stream(
430
 
                        [('D',), ('C',), ('E',), ('F',)], 'unordered', False)]
431
 
        self.assertEqual([(('E',), 'knit-delta-gz'), (('D',), 'knit-delta-gz'),
432
 
                          (('F',), 'knit-delta-gz'), (('C',), 'knit-delta-gz')],
433
 
                          records)
434
 
 
435
 
        target_branch.lock_write()
436
 
        self.addCleanup(target_branch.unlock)
437
 
        target = target_branch.repository
438
 
        target.fetch(source, revision_id='F')
439
 
        # 'C' should be expanded to a fulltext, but D and E should still be
440
 
        # deltas
441
 
        stream = target.inventories.get_record_stream(
442
 
            [('C',), ('D',), ('E',), ('F',)],
443
 
            'unordered', False)
444
 
        kinds = dict((record.key, record.storage_kind) for record in stream)
445
 
        self.assertEqual({('C',): 'knit-ft-gz', ('D',): 'knit-delta-gz',
446
 
                          ('E',): 'knit-delta-gz', ('F',): 'knit-delta-gz'},
447
 
                         kinds)
448
 
 
449
435
 
450
436
class Test1To2Fetch(TestCaseWithTransport):
451
437
    """Tests for Model1To2 failure modes"""
496
482
        self.repo.fetch(self.tree.branch.repository, 'second-id')
497
483
        root_id = self.tree.get_root_id()
498
484
        self.assertEqual(
499
 
            ((root_id, 'left-parent'), (root_id, 'not-ghost-parent')),
 
485
            ((root_id, 'left-parent'), (root_id, 'ghost-parent'),
 
486
             (root_id, 'not-ghost-parent')),
500
487
            self.get_parents(root_id, 'second-id'))
501
488
 
502
489
    def make_two_commits(self, change_root, fetch_twice):