5
5
.. contents:: List of Releases
12
:2.3b1: NOT RELEASED YET
17
* `FileInWrongBranch` is deprecated in favour of `PathNotChild` and no
27
* `PathNotChild` should not give a traceback.
50
:2.2rc1: NOT RELEASED YET
61
* Configuration files in ``${BZR_HOME}`` are now written in an atomic
62
way which should help avoid problems with concurrent writers.
63
(Vincent Ladeuil, #525571)
65
* Don't traceback trying to unversion children files of an already
66
unversioned directory. (Vincent Ladeuil, #494221)
87
:Codename: Monkey Magic
91
This fourth and final beta in the 2.2 series now stabilizes the internal
92
APIs. Plugin authors are recommended to ensure their releases are
93
compatible, so that 2.2rc1 can be a true release candidate, containing
94
stable and compatible plugin versions.
96
For users of bzrlib as a library, one of the primary changes is to request
97
that they call ``bzrlib.initialize`` and use the returned context manager
100
Better interaction with ``bzr-loom`` to make sure branching from a loom
101
even over a smart server still yields a local loom. Not to mention lots of
107
* bzrlib library users now need to call ``__enter__`` and ``__exit__`` on
108
the result of ``bzrlib.initialize``. This change was made when fixing
109
the bad habit recent bzr versions have had of leaving progress bars
110
behind on the screen. That required calling another function before
111
exiting the program, and it made sense to provide a full context
112
manager at the same time. (Robert Collins)
114
* The ``bzr`` front end now requires a ``bzrlib.ui.ui_factory`` which is a
115
context manager in the Python 2.5 and above sense. The bzrlib base class
116
is such a manager, but third party UI factories which do not derive from
117
``bzrlib.ui.UIFactory`` will be incompatible with the command line front
120
* URLs like ``foo:bar/baz`` are now always parsed as a URL with scheme "foo"
121
and path "bar/baz", even if bzr does not recognize "foo" as a known URL
122
scheme. Previously these URLs would be treated as local paths.
129
* Support ``--directory`` option for a number of additional commands:
130
conflicts, merge-directive, missing, resolve, shelve, switch,
131
unshelve, whoami. (Martin von Gagern, #527878)
136
* ``bzr branch`` to a new repository with a default stacking policy no
137
longer transfers the full history unnecessarily.
138
(Andrew Bennetts, #597942)
140
* ``bzr init`` does not recursively scan directory contents anymore
141
leading to faster init for directories with existing content.
142
(Martin [gz], Parth Malwankar, #501307)
144
* ``bzr log --exclude-common-ancestry`` is now taken into account for
145
linear ancetries. (Vincent Ladeuil, #575631)
147
* ``bzr log -r branch:REMOTE`` can now properly log the remote branch,
148
rather than trying to fetch the data locally and failing because of a
149
readonly error. (Martin von Gagern, #149270)
151
* ``bzr pull`` now works when a lp: URL is explicitly defined as the parent
152
or pull location in locations.conf or branch.conf.
153
(Gordon Tyler, #534787)
155
* ``bzr reconfigure --unstacked`` now works with branches accessed via a
156
smart server. (Andrew Bennetts, #551525)
158
* ``BzrDir.find_branches`` should ignore branches with missing repositories.
159
(Marius Kruger, Robert Collins)
161
* ``BzrDir.find_bzrdirs`` should ignore dirs that raises PermissionDenied.
162
(Marius Kruger, Robert Collins)
164
* Ensure that wrong path specifications in ``BZR_PLUGINS_AT`` display
165
proper error messages. (Vincent Ladeuil, #591215)
167
* Explicitly removing ``--profile-imports`` option from parsed command-line
168
arguments on Windows, because bzr script does the same.
169
(Alexander Belchenko, #588277)
171
* Fetching was slightly confused about the best code to use and was
172
using a new code path for all branches, resulting in more lookups than
173
necessary on old branches. (Robert Collins, #593515)
175
* Final fix for 'no help for command' issue. We now show a clean message
176
when a command has no help, document how to set help more clearly, and
177
test that all commands available to the test suite have help.
178
(Robert Collins, #177500)
180
* Invalid patterns supplied to ``Globster`` or ``lazy_regex`` now raise
181
``InvalidPattern`` exception showing clear error message to the user.
182
(Parth Malwankar #300062)
184
* Progress output is cleaned up when exiting. (Aaron Bentley)
186
* Raise ValueError instead of a string exception.
187
(John Arbash Meinel, #586926)
189
* Relative imports in plugins are now handled correctly when using
190
BZR_PLUGINS_AT. (Vincent Ladeuil, #588959)
192
* ``ScriptRunner`` now strips off leading indentation from test scripts,
193
which previously caused "SyntaxError: No command for line".
196
* Show unicode filenames in diff headers using terminal encoding.
197
(Alexander Belchenko, Bug #382699)
198
NOTE for Windows users: If user need to save diff to file then user need to
199
change encoding of the terminal to ANSI encoding with command ``chcp XXX``
200
(e.g. ``chcp 1251`` for Russian Windows).
202
* URL displayed for use with ``break-lock`` when smart server sees lock
203
contention are now valid. Default timeout for lock contention retry is
204
now 30 seconds instead of 300 seconds.
205
(Parth Malwankar, #250451)
207
* ``walkdirs`` now raises a useful message when the filenames are not using
208
the filesystem encoding. (Eric Moritz, #488519)
210
* Enable debugging of bzr on windows with pdb and other tools. This was
211
broken because we call GetCommandLineW on windows. The fix adjusts the
212
command line we get to be the same length as sys.argv.
213
(Jason Spashett, Alexander Belchenko, #587868)
218
* Bazaar now reads data from SSH connections more efficiently on platforms
219
that provide the ``socketpair`` function, and when using paramiko.
220
(Andrew Bennetts, #590637)
222
* ``Branch.copy_content_into`` is now a convenience method dispatching to
223
a ``InterBranch`` multi-method. This permits ``bzr-loom`` and other
224
plugins to intercept this even when a ``RemoteBranch`` proxy is in use.
225
(Robert Collins, #201613)
227
* ``Branch`` formats can now be loaded lazily by registering a
228
``MetaDirBranchFormatFactory`` rather than an actual format. This will
229
cause the named format class to be loaded only when an enumeration of
230
formats is needed or when the format string for the object is
231
encountered. (Robert Collins, Jelmer Vernooij)
233
* The encoding that bzr uses to output things other than file content can
234
now be overridden via the output_encoding configuration option.
235
(Martin Pool, #340394)
237
* Use lazy imports in ``bzrlib/merge.py`` so that plugins like ``news_merge``
238
do not cause modules to be loaded unnecessarily just because the plugin
239
registers a merge hook. This improves ``bzr rocks`` time by about 25%
240
in a default installation (with just the core plugins).
246
* Added ``regression`` tag to our tags list. (Robert Collins)
248
* Improved our release checklist to have a bit less churn and leave things
249
ready-to-go for the next action (including other people doing
250
development). (Robert Collins)
252
* Remove obsolete discussion of PQM in documentation about how to
253
contribute to Bazaar. (Martin Pool, #588444)
258
* ``bzrlib.branch.InterBranch._get_branch_formats_to_test`` now returns
259
an iterable of format pairs, rather than just a single pair, permitting
260
InterBranch objects that work with multiple permutations to be
261
comprehensively tested. (Robert Collins)
263
* ``bzrlib.lsprof.profile`` will no longer silently generate bad threaded
264
profiles when concurrent profile requests are made. Instead the profile
265
requests will be serialised. Reentrant requests will now deadlock.
268
* ``bzrlib.knit.KnitSequenceMatcher``, which has been deprecated since
269
2007, has been deleted. Use ``PatienceSequenceMatcher`` from
270
``bzrlib.patiencediff`` instead. (Andrew Bennetts)
272
* ``bzrlib.re_compile_checked`` is now deprecated. Caller should handle
273
``bzrlib.errors.InvalidPattern`` exception thrown by ``re.match`` in
274
case the default error message not suitable for the use case.
277
* ``bzrlib.tests.blackbox.ExternalBase`` is deprecated. It provided only
278
one method ``check_output``, and we now recommend checking command
279
output using ``run_script``. (Martin Pool)
281
* ``bzrlib.transport.ssh.SSHVendor.connect_ssh`` now returns an object
282
that implements the interface of ``bzrlib.transport.ssh.SSHConnection``.
283
Third-party implementations of ``SSHVendor`` may need to be updated
284
accordingly. Similarly, any code using ``SSHConnection`` directly will
285
need to be updated. (Andrew Bennetts)
287
* The constructor of ``bzrilb.smart.medium.SmartSSHClientMedium`` has
288
changed to take an ``SSHParams`` instance (replacing many individual
289
values). (Andrew Bennetts)
294
* ``bzrlib.osutils.get_terminal_encoding`` will now only mutter its
295
selection when explicitly requested; this avoids many duplicate calls
296
being logged when helpers, wrappers and older code that manually calls
297
it are executed it is now logged deliberately by the ui setup code.
300
* Improved ``bzrlib.urlutils`` to handle lp:foo/bar URLs. (Gordon Tyler)
302
* ``bzrlib._c_static_tuple.StaticTuple`` now implements ``__sizeof__``, so
303
that ``sys.getsizeof`` and other memory analysis tools will report more
304
accurate results. (Andrew Bennetts)
306
* The symbol_versioning module can now cleanup after itself -
307
``suppress_deprecation_warnings`` now returns a cleanup function.
313
* Add ``bzrlib.tests.fixtures`` to hold code for setting up objects
314
to test. (Martin Pool)
316
* ``test_import_tariff`` now respects BZR_PLUGINS_AT and BZR_PLUGINS_DISABLE.
317
(Vincent Ladeuil, #595587)
11
:2.2b3: NOT RELEASED YET
324
This third beta in the 2.2 series brings with it all the goodness of 2.1.2
325
and 2.0.6 (though it preceeds 2.0.6 slightly). Of particular note for
326
users are compatibility fixes with bzr 1.5 and below servers, a hopeful
327
end to the EINTR errors caused by SIGWINCH interactions, a shiny new
328
bash completion script and bzr will no longer guess at identity details -
329
it was too unreliable in reality. Use ``bzr whoami`` on every new install.
330
For developers we have some API changes which may impact plugins as well
331
as a bunch of our regular improvements to internal clarity and test
13
334
Compatibility Breaks
14
335
********************
337
* An API break has been made to the lock_write method of ``Branch`` and
338
``Repository`` objects; they now return ``branch.BranchWriteLockResult``
339
and ``repository.RepositoryWriteLockResult`` objects. This makes
340
changing the API in future easier and permits some cleaner calling code.
341
The lock_read method has also changed from having no defined return
342
value to returning ``LogicalLockResult`` objects.
345
* ``bzr`` does not try to guess the username as ``username@hostname``
346
and requires it to be explictly set. This can be set using ``bzr
347
whoami``. (Parth Malwankar, #549310)
349
* ``bzrlib.commands.Command`` will now raise ValueError during
350
construction if there is no __doc__ set. (Note, this will be reverted in
351
2.2b4) (Robert Collins)
353
* The source tree no longer contains a contrib/zsh/_bzr completion
354
script. The new file contrib/zsh/README suggests alternatives.
355
(Martin von Gagern, #560030)
31
372
(Parth Malwankar, #304320)
374
* New command line option ``--authors`` to ``bzr log`` allows users to
375
select which of the apparent authors and committer should be
376
included in the log. Defaults depend on format. (Martin von Gagern, #513322)
378
* Support ``--directory`` option for a number of additional commands:
379
added, annotate, bind, cat, cat-revision, clean-tree, deleted,
380
export, ignore, ignored, lookup-revision, ls, modified, nick,
381
re-sign, unbind, unknowns.
382
(Martin von Gagern, #527878)
384
* The bash_completion plugin from the bzr-bash-completion project has
385
been merged into the tree. It provides a bash-completion command and
386
replaces the outdated ``contrib/bash/bzr`` script with a version
387
using the plugin. (Martin von Gagern, #560030)
389
* A new transport based on GIO (the gnome i/o library) provides access to
390
samba shares, webdav using gio+smb and gio+dav. It is also possible to
391
use gio for some already existing transport methods as gio+file,
398
* Alias information shown by ``bzr help`` is now accurate. This
399
was showing an internal object name for some plugin aliases.
400
(Parth Malwankar, #584650)
36
402
* ``.bazaar``, ``.bazaar/bazaar.conf`` and ``.bzr.log`` inherit user and
37
403
group ownership from the containing directory. This allow bzr to work
39
405
(Martin <gzlist@googlemail.com>, Parth Malwankar, #376388)
407
* ``bzr clean-tree`` should not delete nested bzrdirs. Required for proper
408
support of bzr-externals and scmproj plugins.
409
(Alexander Belchenko, bug #572098)
411
* ``bzr ignore`` will no longer add duplicate patterns to .bzrignore.
412
(Gordon Tyler, #572092)
41
414
* ``bzr log --exclude-common-ancestry -r X..Y`` displays the revisions that
42
415
are part of Y ancestry but not part of X ancestry (aka the graph
44
417
(Vincent Ladeuil, #320119)
419
* ``bzr lp-propose`` which was switched to use production Launchpad API
420
servers a few commits ago has been reverted to use edge: there is a
421
problem with using production which isn't trivially obvious, so we've
422
filed a bug to track it, and until thats fixed will be using edge.
423
(Robert Collins, #583667)
425
* ``bzr rm`` should not refuse to delete directories which contained a file
426
which has been moved elsewhere in the tree after the previous commit.
427
(Marius Kruger, Daniel Watkins, #129880)
46
429
* ``bzr selftest --parallel=fork`` wait for its children avoiding zombies.
47
430
(Vincent Ladeuil, #566670)
49
432
* ``bzr selftest`` should not use ui.note() since it's not unicode safe.
50
433
(Vincent Ladeuil, #563997)
435
* CommitBuilder refuses to create revisions whose trees have no root.
438
* Do not register a SIGWINCH signal handler, instead just poll for the
439
terminal width as needed. This avoids the "Interrupted System Call"
440
problems that occur on POSIX with all currently released versions of
442
(Andrew Bennetts, #583941)
52
444
* Don't mention --no-strict when we just issue the warning about unclean trees.
53
445
(Vincent Ladeuil, #401599)
56
448
versions before 1.6.
57
449
(Andrew Bennetts, #528041)
451
* Improved progress bar for fetch (2a format only). Bazaar now shows an
452
estimate of the number of records to be fetched vs actually fetched.
453
(Parth Malwankar, #374740, #538868)
59
455
* Reduce peak memory by one copy of compressed text.
60
456
(John Arbash Meinel, #566940)
458
* ``RemoteBranch.lock_write`` raises ``ReadOnlyError`` if called during a
459
read lock, rather than causing an ``AttributeError``.
460
(Andrew Bennetts, Данило Шеган, #582781)
462
* Selftest was failing with testtools 0.9.3, which caused an
463
AssertionError raised from a cleanUp to be reported as a Failure, not an
464
Error, breaking on of our test hygiene tests.
465
(Robert Collins, Vincent Ladeuil).
467
* ``set_user_option`` with a dict on remote branches no longer fails with
468
an AttributeError. There is a new ``Branch.set_config_option_dict`` RPC
469
to support this efficiently.
470
(Andrew Bennetts, #430382)
62
472
* Show the filenames when a file rename fails so that the error will be
63
473
more comprehensible.
64
474
(Martin Pool, #491763)
476
* Support Pyrex 0.9.9, required changing how we handle exceptions in Pyrex.
477
(John Arbash Meinel, #582656)
479
* Unicode characters in aliases are now handled correctly and do not cause
480
UnicodeEncodeError exception. (Parth Malwankar, #529930)
482
* Unicode commit messages that are the same as a file name no longer cause
483
UnicodeEncodeError. ``ui.text.show_warning`` now handles unicode
485
(Parth Malwankar, #563646)
487
* Using bzr with `lp:` urls behind an http proxy should work.
488
(Robert Collins, #558343)
490
* When passing a file to ``UTF8DirReader`` make sure to close the current
491
directory file handle after the chdir fails. Otherwise when passing many
492
filenames into a command line ``bzr status`` we would leak descriptors.
493
(John Arbash Meinel, #583486)