Log Message:
bzr 1.16.1 2009-06-26
#####################

End user testing of the 2a format revealed two serious bugs. The first,
#365615, caused bzr to raise AbsentContentFactory errors when autopacking.
This meant that commits or pushes to 2a-format repositories failed
intermittently.

The second bug, #390563, caused the smart server to raise AbsentContentFactory
when streaming 2a stacked 2a-format branches. This particularly affected
branches stored on Launchpad in the 2a format.

Bug Fixes
*********

* We now properly request a more minimal set of file texts when fetching
  multiple revisions. (Robert Collins, John Arbash Meinel, #390563)

* Repositories using CHK pages (which includes the new 2a format) will no
  longer error during commit or push operations when an autopack operation
  is triggered. (Robert Collins, #365615)

* ``chk_map.iter_interesting_nodes`` now properly uses the *intersection*
  of referenced nodes rather than the *union* to determine what
  uninteresting pages we still need to look at. Prior to this,
  incrementally pushing to stacked branch would push the minimal data, but
  fetching everything would request extra texts. There are some unhandled
  cases wrt trees of different depths, but this fixes the common cases.
  (Robert Collins, John Arbash Meinel, #390563)

* ``GroupCompress`` repositories now take advantage of the pack hints
  parameter to permit cross-format fetching to incrementally pack the
  converted data. (Robert Collins)

* ``Repository.commit_write_group`` now returns opaque data about what
  was committed, for passing to the ``Repository.pack``. Repositories
  without atomic commits will still return None. (Robert Collins)

* ``Repository.pack`` now takes an optional ``hint`` parameter
  which will support doing partial packs for repositories that can do
  that. (Robert Collins)

* RepositoryFormat has a new attribute 'pack_compresses' which is True
  when doing a pack operation changes the compression of content in the
  repository. (Robert Collins)

* ``StreamSink`` and ``InterDifferingSerialiser`` will call
  ``Repository.pack`` with the hint returned by
  ``Repository.commit_write_group`` if the formats were different and the
  repository can increase compression by doing a pack operation.
  (Robert Collins, #376748)

bzr 1.16 "It's yesterday in California" 2009-06-18
##################################################
:Codename: yesterday-in-california
:1.16rc1: 2009-06-11
:1.16: 2009-06-18

This version of Bazaar contains the beta release of the new ``2a`` repository
format, suitable for testing by fearless, advanced users. This format or an
updated version of it will become the default format in Bazaar 2.0. Please
read the NEWS entry before even thinking about upgrading to the new format.

Also included are speedups for many operations on huge projects, a bug fix for
pushing stacked new stacked branches to smart servers and the usual bevy of
bug fixes and improvements.

Changes from 1.16rc1 to 1.16final
*********************************

* Fix the nested tree flag check so that upgrade from development formats to
  2a can work correctly.
  (Jelmer Vernooij, #388727)

* Automatic format upgrades triggered by default stacking policies on a
  1.16rc1 (or later) smart server work again.
  (Andrew Bennetts, #388675)

Compatibility Breaks
********************

* Display prompt on stderr (instead of stdout) when querying users so
  that the output of commands can be safely redirected.
  (Vincent Ladeuil, #376582)

New Features
************

* A new repository format ``2a`` has been added.  This is a beta release
  of the the brisbane-core (aka group-compress) project.  This format now
  suitable for wider testing by advanced users willing to deal with some
  bugs.  We would appreciate test reports, either positive or negative.
  Format 2a is substantially smaller and faster for many operations on
  many trees.  This format or an updated version will become the default
  in bzr 2.0.

  This is a rich-root format, so this repository format can be used with
  bzr-svn.  Bazaar branches in previous non-rich-root formats can be
  converted (including by merge, push and pull) to format 2a, but not vice
  versa.  We recommend upgrading previous development formats to 2a.

  Upgrading to this format can take considerable time because it expands
  and more concisely repacks the full history.

  If you use stacked branches, you must upgrade the stacked branches
  before the stacked-on branches.  (See \ 
<https://bugs.launchpad.net/bugs/374735>)

* ``--development7-rich-root`` is a new dev format, similar to ``--dev6``
  but using a Revision serializer using bencode rather than XML.
  (Jelmer Vernooij, John Arbash Meinel)

* mail_client=claws now supports --body (and message body hooks).  Also uses
  configured from address.  (Barry Warsaw)

Improvements
************

* ``--development6-rich-root`` can now stack. (Modulo some smart-server
  bugs with stacking and non default formats.)
  (John Arbash Meinel, #373455)

* ``--development6-rich-root`` delays generating a delta index for the
  first object inserted into a group. This has a beneficial impact on
  ``bzr commit`` since each committed texts goes to its own group. For
  committing a 90MB file, it drops peak memory by about 200MB, and speeds
  up commit from 7s => 4s. (John Arbash Meinel)

* Numerous operations are now faster for huge projects, i.e. those
  with a large number of files and/or a large number of revisions,
  particularly when the latest development format is used. These
  operations (and improvements on OpenOffice.org) include:

  * branch in a shared repository (2X faster)
  * branch --no-tree (100X faster)
  * diff (2X faster)
  * tags (70X faster)

  (Ian Clatworthy)

* Pyrex version of ``bencode`` support. This provides optimized support
  for both encoding and decoding, and is now found at ``bzrlib.bencode``.
  ``bzrlib.utils.bencode`` is now deprecated.
  (Alexander Belchenko, Jelmer Vernooij, John Arbash Meinel)

Bug Fixes
*********

* Bazaar can now pass attachment files to the mutt email client.
  (Edwin Grubbs, #384158)

* Better message in ``bzr add`` output suggesting using ``bzr ignored`` to
  see which files can also be added.  (Jason Spashett, #76616)

* ``bzr pull -r 123`` from a stacked branch on a smart server no longer fails.
  Also, the ``Branch.revision_history()`` API now works in the same
  situation.  (Andrew Bennetts, #380314)

* ``bzr serve`` on Windows no longer displays a traceback simply because a
  TCP client disconnected. (Andrew Bennetts)

* Clarify the rules for locking and fallback repositories. Fix bugs in how
  ``RemoteRepository`` was handling fallbacks along with the
  ``_real_repository``. (Andrew Bennetts, John Arbash Meinel, #375496)

* Fix a small bug with fetching revisions w/ ghosts into a new stacked
  branch. Not often triggered, because it required ghosts to be part of
  the fetched revisions, not in the stacked-on ancestry.
  (John Arbash Meinel)

* Fix status and commit to work with content filtered trees, addressing
  numerous bad bugs with line-ending support. (Ian Clatworthy, #362030)

* Fix problem of "directory not empty" when contending for a lock over
  sftp.  (Martin Pool, #340352)

* Fix rule handling so that eol is optional, not mandatory.
  (Ian Clatworthy, #379370)

* Pushing a new stacked branch to a 1.15 smart server was broken due to a
  bug in the ``BzrDirFormat.initialize_ex`` smart verb.  This is fixed in
  1.16, but required changes to the network protocol, so the
  ``BzrDirFormat.initialize_ex`` verb has been removed and replaced with a
  corrected ``BzrDirFormat.initialize_ex_1.16`` verb.  1.15 clients will
  still work with a 1.16 server as they will fallback to slower (and
  bug-free) methods.
  (Jonathan Lange, Robert Collins, Andrew Bennetts, #385132)

* Reconcile can now deal with text revisions that originated in revisions
  that are ghosts. (Jelmer Vernooij, #336749)

* Support cloning of branches with ghosts in the left hand side history.
  (Jelmer Vernooij, #248540)

* The ''bzr diff'' now catches OSError from osutils.rmtree and logs a
  helpful message to the trace file, unless the temp directory really was
  removed (which would be very strange).  Since the diff operation has
  succeeded from the user's perspective, no output is written to stderr
  or stdout.  (Maritza Mendez, #363837)

* Translate errors received from a smart server in response to a
  ``BzrDirFormat.initialize`` or ``BzrDirFormat.initialize_ex`` request.
  This was causing tracebacks even for mundane errors like
  ``PermissionDenied``.  (Andrew Bennetts, #381329)

Documentation
*************

* Added directory structure and started translation of docs in Russian.
  (Alexey Shtokalo, Alexander Iljin, Alexander Belchenko, Dmitry Vasiliev,
  Volodymyr Kotulskyi)

API Changes
***********

* Added osutils.parent_directories(). (Ian Clatworthy)

* ``bzrlib.progress.ProgressBar``, ``ChildProgress``, ``DotsProgressBar``,
  ``TTYProgressBar`` and ``child_progress`` are now deprecated; use
  ``ui_factory.nested_progress_bar`` instead.  (Martin Pool)

* ``graph.StackedParentsProvider`` is now a public API, replacing
  ``graph._StackedParentsProvider``. The api is now considered stable and ready
  for external users. (Gary van der Merwe)

* ``bzrlib.user_encoding`` is deprecated in favor of
  ``get_user_encoding``.  (Alexander Belchenko)

* TreeTransformBase no longer assumes that limbo is provided via disk.
  DiskTreeTransform now provides disk functionality.  (Aaron Bentley)

Internals
*********

* Remove ``weave.py`` script for accessing internals of old weave-format
  repositories.  (Martin Pool)

Testing
*******

* The number of cores is now correctly detected on OSX. (John Szakmeister)

* The number of cores is also detected on Solaris and win32. (Vincent Ladeuil)

* The number of cores is also detected on FreeBSD. (Matthew Fuller)

bzr 1.15
########
:1.15rc1: 2009-05-16
:1.15: 2009-05-22
:1.15.1: 2009-06-09

The smart server will no longer raise 'NoSuchRevision' when streaming content
with a size mismatch in a reconstructed graph search. New command ``bzr
dpush``. Plugins can now define their own annotation tie-breaker when two
revisions introduce the exact same line.

Changes from 1.15.1 to 1.15.2
*****************************

* Use zdll on Windows to build ``_chk_map_pyx`` extension.
  (Alexander Belchenko)

Changes from 1.15final to 1.15.1
*********************************

* Translate errors received from a smart server in response to a
  ``BzrDirFormat.initialize`` or ``BzrDirFormat.initialize_ex`` request.
  This was causing tracebacks even for mundane errors like
  ``PermissionDenied``.  (Andrew Bennetts, #381329)

Changes from 1.15rc1 to 1.15final
*********************************

* No changes

Compatibility Breaks
********************

* ``bzr ls`` is no longer recursive by default. To recurse, use the
  new ``-R`` option. The old ``--non-recursive`` option has been removed.
  If you alias ``ls`` to ``ls -R``, you can disable recursion using
  ``--no-recursive`` instead.  (Ian Clatworthy)

New Features
************

* New command ``bzr dpush`` that can push changes to foreign
  branches (svn, git) without setting custom bzr-specific metadata.
  (Jelmer Vernooij)

* The new development format ``--development6-rich-root`` now supports
  stacking. We chose not to use a new format marker, since old clients
  will just fail to open stacked branches, the same as if we used a new
  format flag. (John Arbash Meinel, #373455)

* Plugins can now define their own annotation tie-breaker when two revisions
  introduce the exact same line. See ``bzrlib.annotate._break_annotation_tie``
  Be aware though that this is temporary, private (as indicated by the leading
  '_') and a first step to address the problem. (Vincent Ladeuil, #348459)

* New command ``bzr dpush`` that can push changes to foreign
  branches (svn, git) without setting custom bzr-specific metadata.
  (Jelmer Vernooij)

* ``bzr send`` will now check the ``child_submit_format`` setting in
  the submit branch to determine what format to use, if none was
  specified on the command-line.  (Jelmer Vernooij)

Improvements
************

* -Dhpss output now includes the number of VFS calls made to the remote
  server. (Jonathan Lange)

* ``--coverage`` works for code running in threads too.
  (Andrew Bennets, Vincent Ladeuil)

* ``bzr pull`` now has a ``--local`` option to only make changes to the
  local branch, and not the bound master branch.
  (Gary van der Merwe, #194716)

* ``bzr rm *`` is now as fast as ``bzr rm * --keep``. (Johan Walles, #180116)

Bug Fixes
*********

* Adding now works properly when path contains a symbolic link.
  (Geoff Bache, #183831)

* An error is now raised for unknown eol values. (Brian de Alwis, #358199)

* ``bzr merge --weave`` will now generate a conflict if one side deletes a
  line, and the other side modifies the line. (John Arbash Meinel, #328171)

* ``bzr reconfigure --standalone`` no longer raises IncompatibleRepositories.
  (Martin von Gagern, #248932)

* ``bzr send`` works to send emails again using MAPI.
  (Neil Martinsen-Burrell, #346998)

* Check for missing parent inventories in StreamSink.  This prevents
  incomplete stacked branches being created by 1.13 bzr:// and
  bzr+ssh:// clients (which have bug #354036).  Instead, the server now
  causes those clients to send the missing records.  (Andrew Bennetts)

* Correctly handle http servers proposing multiple authentication schemes.
  (Vincent Ladeuil, #366107)

* End-Of-Line content filters are now loaded correctly.
  (Ian Clatworthy, Brian de Alwis, #355280)

* Fix a bug in the pure-python ``GroupCompress`` code when handling copies
  longer than 64KiB. (John Arbash Meinel, #364900)

* Fix TypeError in running ``bzr break-lock`` on some URLs.
  (Alexander Belchenko, Martin Pool, #365891)

* Non-recursive ``bzr ls`` now works properly when a path is specified.
  (Jelmer Vernooij, #357863)

* ssh usernames (defined in ~/.ssh/config) are honoured for bzr+ssh connections.
  (Vincent Ladeuil, #367726)

* Several bugs related to unicode symlinks have been fixed and the test suite
  enhanced to better catch regressions for them. (Vincent Ladeuil)

* The smart server will no longer raise 'NoSuchRevision' when streaming
  content with a size mismatch in a reconstructed graph search: it assumes
  that the client will make sure it is happy with what it got, and this
  sort of mismatch is normal for stacked environments.
  bzr 1.13.0/1 will stream from unstacked branches only - in that case not
  getting all the content expected would be a bug. However the graph
  search is how we figured out what we wanted, so a mismatch is both odd
  and unrecoverable without starting over, and starting over will end up
  with the same data as if we just permitted the mismatch. If data is
  gc'd, doing a new search will find only the truncated data, so sending
  only the truncated data seems reasonable. bzr versions newer than this
  will stream from stacked branches and check the stream to find missing
  content in the stacked-on branch, and thus will handle the situation
  implicitly.  (Robert Collins, #360791)

* Upgrading to, or fetching into a 'rich-root' format will now correctly
  set the root data the same way that reconcile does.
  (Robert Collins, #368921)

* Using unicode Windows API to obtain command-line arguments.
  (Alexander Belchenko, #375934)

Documentation
*************

API Changes
***********

* ``InterPackRepo.fetch`` and ``RepoFetcher`` now raise ``NoSuchRevision``
  instead of ``InstallFailed`` when they detect a missing revision.
  ``InstallFailed`` itself has been deleted. (Jonathan Lange)

* Not passing arguments to ``bzrlib.commands.main()`` will now grab the
  arguments from ``osutils.get_unicode_argv()`` which has proper support
  for unicode arguments on windows. Further, the supplied arguments are now
  required to be unicode strings, rather than user_encoded strings.
  (Alexander Belchenko)

Internals
*********

* ``bzrlib.branch.Branch.set_parent`` is now present on the base branch
  class and will call ``_set_parent_location`` after doing unicode
  encoding. (Robert Collins)

* ``bzrlib.remote.RemoteBranch._set_parent_location`` will use a new verb
  ``Branch.set_parent_location`` removing further VFS operations.
  (Robert Collins)

* ``bzrlib.bzrdir.BzrDir._get_config`` now returns a ``TransportConfig``
  or similar when the dir supports configuration settings. The base class
  defaults to None. There is a matching new server verb
  ``BzrDir.get-config_file`` to reduce roundtrips for getting BzrDir
  configuration. (Robert Collins)

* ``bzrlib.tests.ExtendedTestResult`` has new methods ``startTests``
  called before the first test is started, ``done`` called after the last
  test completes, and a new parameter ``strict``. (Robert Collins)

* ``-Dhpss`` when passed to bzr will cause a backtrace to be printed when
  VFS operations are started on a smart server repository. This should not
  occur on regular push and pull operations, and is a key indicator for
  performance regressions. (Robert Collins)

* ``-Dlock`` when passed to the selftest (e.g. ``bzr -Dlock selftest``) will
  cause mismatched physical locks to cause test errors rather than just
  reporting to the screen. (Robert Collins)

* Fallback ``CredentialStore`` instances registered with ``fallback=True``
  are now be able to provide credentials if obtaining credentials
  via ~/.bazaar/authentication.conf fails. (Jelmer Vernooij,
  Vincent Ladeuil, #321918)

* New hook ``Lock.lock_broken`` which runs when a lock is
  broken. This is mainly for testing that lock/unlock are
  balanced in tests. (Vincent Ladeuil)

* New MergeDirective hook 'merge_request_body' allows hooks to supply or
  alter a body for the message produced by ``bzr send``.

* New smart server verb ``BzrDir.initialize_ex`` which implements a
  refactoring to the core of clone allowing less round trips on new
  branches. (Robert Collins)

* New method ``Tags.rename_revisions`` that can rename revision ids tags
  are pointing at. (Jelmer Vernooij)

* Updated the bundled ``ConfigObj`` library to 4.6.0 (Matt Nordhoff)

Testing
*******

* ``bzr selftest`` will now fail if lock/unlock are not correctly balanced in
  tests. Using ``-Dlock`` will turn the related failures into warnings.
  (Vincent Ladeuil, Robert Collins)