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
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
22
from bzrlib import (
21
revision as _mod_revision,
24
28
from bzrlib.branch import Branch
29
from bzrlib.bzrdir import BzrDir
25
30
from bzrlib.repofmt import knitrepo
26
31
from bzrlib.tests import TestCaseWithTransport
32
from bzrlib.tests.http_utils import TestCaseWithWebserver
27
33
from bzrlib.tests.test_revision import make_branches
34
from bzrlib.trace import mutter
28
35
from bzrlib.upgrade import Convert
29
36
from bzrlib.workingtree import WorkingTree
31
38
# These tests are a bit old; please instead add new tests into
32
# per_interrepository/ so they'll run on all relevant
39
# interrepository_implementations/ so they'll run on all relevant
36
43
def has_revision(branch, revision_id):
37
44
return branch.repository.has_revision(revision_id)
40
def revision_history(branch):
43
graph = branch.repository.get_graph()
44
history = list(graph.iter_lefthand_ancestry(branch.last_revision(),
45
[_mod_revision.NULL_REVISION]))
52
46
def fetch_steps(self, br_a, br_b, writable_a):
53
47
"""A foreign test method for testing fetch locally and remotely."""
55
49
# TODO RBC 20060201 make this a repository test.
56
50
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])
51
self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
52
self.assertTrue(repo_b.has_revision(br_a.revision_history()[2]))
53
self.assertEquals(len(br_b.revision_history()), 7)
54
self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[2])[0], 0)
61
55
# 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]))
56
self.assertEquals(len(br_b.revision_history()), 7)
57
self.assertFalse(repo_b.has_revision(br_a.revision_history()[3]))
65
59
# 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]))
60
self.assertEquals(br_b.fetch(br_a, br_a.revision_history()[3])[0], 1)
61
self.assertTrue(repo_b.has_revision(br_a.revision_history()[3]))
62
self.assertFalse(has_revision(br_a, br_b.revision_history()[6]))
63
self.assertTrue(br_a.repository.has_revision(br_b.revision_history()[5]))
71
65
# When a non-branch ancestor is missing, it should be unlisted...
72
66
# as its not reference from the inventory weave.
73
67
br_b4 = self.make_branch('br_4')
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]))
68
count, failures = br_b4.fetch(br_b)
69
self.assertEqual(count, 7)
70
self.assertEqual(failures, [])
72
self.assertEqual(writable_a.fetch(br_b)[0], 1)
73
self.assertTrue(has_revision(br_a, br_b.revision_history()[3]))
74
self.assertTrue(has_revision(br_a, br_b.revision_history()[4]))
80
76
br_b2 = self.make_branch('br_b2')
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]))
77
self.assertEquals(br_b2.fetch(br_b)[0], 7)
78
self.assertTrue(has_revision(br_b2, br_b.revision_history()[4]))
79
self.assertTrue(has_revision(br_b2, br_a.revision_history()[2]))
80
self.assertFalse(has_revision(br_b2, br_a.revision_history()[3]))
86
82
br_a2 = self.make_branch('br_a2')
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]))
83
self.assertEquals(br_a2.fetch(br_a)[0], 9)
84
self.assertTrue(has_revision(br_a2, br_b.revision_history()[4]))
85
self.assertTrue(has_revision(br_a2, br_a.revision_history()[3]))
86
self.assertTrue(has_revision(br_a2, br_a.revision_history()[2]))
92
88
br_a3 = self.make_branch('br_a3')
93
# pulling a branch with no revisions grabs nothing, regardless of
89
# pulling a branch with no revisions grabs nothing, regardless of
94
90
# whats in the inventory.
91
self.assertEquals(br_a3.fetch(br_a2)[0], 0)
96
92
for revno in range(4):
98
br_a3.repository.has_revision(revision_history(br_a)[revno]))
99
br_a3.fetch(br_a2, revision_history(br_a)[2])
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)
100
96
# 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
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
103
100
# that was requested.
104
self.assertRaises(errors.NoSuchRevision, br_a3.fetch, br_a2, 'pizza')
101
self.assertRaises(errors.InstallFailed, br_a3.fetch, br_a2, 'pizza')
106
103
# TODO: Test trying to fetch from a branch that points to a revision not
107
104
# actually present in its repository. Not every branch format allows you
252
249
rev_id).get_file_text('this-file-id'), text)
255
class TestKnitToPackFetch(TestCaseWithTransport):
257
def find_get_record_stream(self, calls, expected_count=1):
258
"""In a list of calls, find the last 'get_record_stream'.
260
:param expected_count: The number of calls we should exepect to find.
261
If a different number is found, an assertion is raised.
263
get_record_call = None
266
if call[0] == 'get_record_stream':
268
get_record_call = call
269
self.assertEqual(expected_count, call_count)
270
return get_record_call
272
def test_fetch_with_deltas_no_delta_closure(self):
273
tree = self.make_branch_and_tree('source', format='dirstate')
274
target = self.make_repository('target', format='pack-0.92')
275
self.build_tree(['source/file'])
276
tree.set_root_id('root-id')
277
tree.add('file', 'file-id')
278
tree.commit('one', rev_id='rev-one')
279
source = tree.branch.repository
280
source.texts = versionedfile.RecordingVersionedFilesDecorator(
282
source.signatures = versionedfile.RecordingVersionedFilesDecorator(
284
source.revisions = versionedfile.RecordingVersionedFilesDecorator(
286
source.inventories = versionedfile.RecordingVersionedFilesDecorator(
289
self.assertTrue(target._format._fetch_uses_deltas)
290
target.fetch(source, revision_id='rev-one')
291
self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
292
target._format._fetch_order, False),
293
self.find_get_record_stream(source.texts.calls))
294
self.assertEqual(('get_record_stream', [('rev-one',)],
295
target._format._fetch_order, False),
296
self.find_get_record_stream(source.inventories.calls, 2))
297
self.assertEqual(('get_record_stream', [('rev-one',)],
298
target._format._fetch_order, False),
299
self.find_get_record_stream(source.revisions.calls))
300
# XXX: Signatures is special, and slightly broken. The
301
# standard item_keys_introduced_by actually does a lookup for every
302
# signature to see if it exists, rather than waiting to do them all at
303
# once at the end. The fetch code then does an all-at-once and just
304
# allows for some of them to be missing.
305
# So we know there will be extra calls, but the *last* one is the one
307
signature_calls = source.signatures.calls[-1:]
308
self.assertEqual(('get_record_stream', [('rev-one',)],
309
target._format._fetch_order, False),
310
self.find_get_record_stream(signature_calls))
312
def test_fetch_no_deltas_with_delta_closure(self):
313
tree = self.make_branch_and_tree('source', format='dirstate')
314
target = self.make_repository('target', format='pack-0.92')
315
self.build_tree(['source/file'])
316
tree.set_root_id('root-id')
317
tree.add('file', 'file-id')
318
tree.commit('one', rev_id='rev-one')
319
source = tree.branch.repository
320
source.texts = versionedfile.RecordingVersionedFilesDecorator(
322
source.signatures = versionedfile.RecordingVersionedFilesDecorator(
324
source.revisions = versionedfile.RecordingVersionedFilesDecorator(
326
source.inventories = versionedfile.RecordingVersionedFilesDecorator(
328
# XXX: This won't work in general, but for the dirstate format it does.
329
self.overrideAttr(target._format, '_fetch_uses_deltas', False)
330
target.fetch(source, revision_id='rev-one')
331
self.assertEqual(('get_record_stream', [('file-id', 'rev-one')],
332
target._format._fetch_order, True),
333
self.find_get_record_stream(source.texts.calls))
334
self.assertEqual(('get_record_stream', [('rev-one',)],
335
target._format._fetch_order, True),
336
self.find_get_record_stream(source.inventories.calls, 2))
337
self.assertEqual(('get_record_stream', [('rev-one',)],
338
target._format._fetch_order, True),
339
self.find_get_record_stream(source.revisions.calls))
340
# XXX: Signatures is special, and slightly broken. The
341
# standard item_keys_introduced_by actually does a lookup for every
342
# signature to see if it exists, rather than waiting to do them all at
343
# once at the end. The fetch code then does an all-at-once and just
344
# allows for some of them to be missing.
345
# So we know there will be extra calls, but the *last* one is the one
347
signature_calls = source.signatures.calls[-1:]
348
self.assertEqual(('get_record_stream', [('rev-one',)],
349
target._format._fetch_order, True),
350
self.find_get_record_stream(signature_calls))
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
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
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
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)
382
def test_fetch_with_fallback_and_merge(self):
383
builder = self.make_branch_builder('source', format='pack-0.92')
384
builder.start_series()
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.
403
('add', ('', 'TREE_ROOT', 'directory', None))]
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
421
self.addCleanup(source.unlock)
422
source.inventories = versionedfile.OrderingVersionedFilesDecorator(
424
key_priority={('E',): 1, ('D',): 2, ('C',): 4,
426
# Ensure that the content is yielded in the proper order, and given as
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')],
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
441
stream = target.inventories.get_record_stream(
442
[('C',), ('D',), ('E',), ('F',)],
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'},
252
class TestHttpFetch(TestCaseWithWebserver):
253
# FIXME RBC 20060124 this really isn't web specific, perhaps an
254
# instrumented readonly transport? Can we do an instrumented
255
# adapter and use self.get_readonly_url ?
257
def test_fetch(self):
258
#highest indices a: 5, b: 7
259
br_a, br_b = make_branches(self)
260
br_rem_a = Branch.open(self.get_readonly_url('branch1'))
261
fetch_steps(self, br_rem_a, br_b, br_a)
263
def _count_log_matches(self, target, logs):
264
"""Count the number of times the target file pattern was fetched in an http log"""
265
get_succeeds_re = re.compile(
266
'.*"GET .*%s HTTP/1.1" 20[06] - "-" "bzr/%s' %
267
( target, bzrlib.__version__))
270
if get_succeeds_re.match(line):
274
def test_weaves_are_retrieved_once(self):
275
self.build_tree(("source/", "source/file", "target/"))
276
# This test depends on knit dasta storage.
277
wt = self.make_branch_and_tree('source', format='dirstate-tags')
279
wt.add(["file"], ["id"])
280
wt.commit("added file")
281
open("source/file", 'w').write("blah\n")
282
wt.commit("changed file")
283
target = BzrDir.create_branch_and_repo("target/")
284
source = Branch.open(self.get_readonly_url("source/"))
285
self.assertEqual(target.fetch(source), (2, []))
286
# this is the path to the literal file. As format changes
287
# occur it needs to be updated. FIXME: ask the store for the
289
self.log("web server logs are:")
290
http_logs = self.get_readonly_server().logs
291
self.log('\n'.join(http_logs))
292
# unfortunately this log entry is branch format specific. We could
293
# factor out the 'what files does this format use' to a method on the
294
# repository, which would let us to this generically. RBC 20060419
295
# RBC 20080408: Or perhaps we can assert that no files are fully read
297
self.assertEqual(1, self._count_log_matches('/ce/id.kndx', http_logs))
298
self.assertEqual(1, self._count_log_matches('/ce/id.knit', http_logs))
299
self.assertEqual(1, self._count_log_matches('inventory.kndx', http_logs))
300
# this r-h check test will prevent regressions, but it currently already
301
# passes, before the patch to cache-rh is applied :[
302
self.assertTrue(1 >= self._count_log_matches('revision-history',
304
self.assertTrue(1 >= self._count_log_matches('last-revision',
306
# FIXME naughty poking in there.
307
self.get_readonly_server().logs = []
308
# check there is nothing more to fetch. We take care to re-use the
309
# existing transport so that the request logs we're about to examine
310
# aren't cluttered with redundant probes for a smart server.
311
# XXX: Perhaps this further parameterisation: test http with smart
312
# server, and test http without smart server?
313
source = Branch.open(
314
self.get_readonly_url("source/"),
315
possible_transports=[source.bzrdir.root_transport])
316
self.assertEqual(target.fetch(source), (0, []))
317
# should make just two requests
318
http_logs = self.get_readonly_server().logs
319
self.log("web server logs are:")
320
self.log('\n'.join(http_logs))
321
self.assertEqual(1, self._count_log_matches('branch-format', http_logs))
322
self.assertEqual(1, self._count_log_matches('branch/format', http_logs))
323
self.assertEqual(1, self._count_log_matches('repository/format',
325
self.assertTrue(1 >= self._count_log_matches('revision-history',
327
self.assertTrue(1 >= self._count_log_matches('last-revision',
329
self.assertEqual(4, len(http_logs))
450
332
class Test1To2Fetch(TestCaseWithTransport):
added
removed
bzrlib/tests/test_fetch.py
