A couple regressions were found in the 1.6 release. There was a performance issue when using bzr+ssh to branch large repositories, and some problems with stacking and rich-root capable repositories.
BUG FIXES:
- Copying between rich-root and rich-root-pack (and vice versa) was accidentally using the inter-model fetcher, instead of recognizing that both were 'rich root' formats. (John Arbash Meinel, #264321)
This release fixes a few regressions found in the 1.6 client. Fetching changes was using an O(N^2) buffering algorithm, so for large projects it would cause memory thrashing. There is also a specific problem with the --1.6-rich-root format, which prevented stacking on top of --rich-root-pack repositories, and could allow users to accidentally fetch experimental data (-subtree) without representing it properly. The --1.6-rich-root format has been deprecated and users are recommended to upgrade to --1.6.1-rich-root immediately. Also we re-introduced a workaround for users who have repositories with incorrect nodes (not possible if you only used official releases). I should also clarify that none of this is data loss level issues, but still sufficient enough to warrant an updated release.
BUG FIXES:
- RemoteTransport.readv() was being inefficient about how it buffered the readv data and processed it. It would keep appending to the same string (causing many copies) and then pop bytes out of the start of the string (causing more copies). With this patch "bzr+ssh://local" can improve dramatically, especially for projects with large files. (John Arbash Meinel)
- Revision texts were always meant to be stored as fulltexts. There was a bug in a bzr.dev version that would accidentally create deltas when copying from a Pack repo to a Knit repo. This has been fixed, but to support those repositories, we know always request full texts for Revision texts. (John Arbash Meinel, #261339)
- The previous --1.6-rich-root format used an incorrect xml serializer, which would accidentally support fetching from a repository that supported subtrees, even though the local one would not. We deprecated that format, and introduced a new one that uses the correct serializer --1.6.1-rich-root. (John Arbash Meinel, #262333)
Finally, the long awaited bzr 1.6 has been released. This release includes new features like Stacked Branches, improved weave merge, and an updated server protocol (now on v3) which will allow for better cross version compatibility. With this release we have deprecated Knit format repositories, and recommend that users upgrade them, we will continue to support reading and writing them for the forseeable future, but we will not be tuning them for performance as pack repositories have proven to be better at scaling. This will also be the first release to bundle TortoiseBzr in the standalone Windows installer.
BUG FIXES:
- Disable automatic detection of stacking based on a containing directory of the target. It interacted badly with push, and needs a bit more work to get the edges polished before it should happen automatically. (John Arbash Meinel, #259275)
BUG FIXES:
- Fix a regression in knit => pack fetching. We had a logic inversion, causing the fetch to insert fulltexts in random order, rather than preserving deltas. (John Arbash Meinel, #256757)
CHANGES:
- Disable reading .bzrrules as a per-branch rule preferences file. The feature was not quite ready for a full release. (Robert Collins)
IMPROVEMENTS:
- Update the windows installer to bundle TortoiseBzr and qbzr into the standalone installer. This will be the first official windows release that installs Tortoise by default. (Mark Hammond)
BUG FIXES:
- Fix a regression in bzr+http support. There was a missing function (_read_line) that needed to be carried over from bzr+ssh support. (Andrew Bennetts)
- GraphIndex objects will internally read an entire index if more than 1/20th of their keyspace is requested in a single operation. This largely mitigates a performance regression in bzr log FILE and completely corrects the performance regression in bzr log. The regression was caused by removing an accomodation which had been supporting the index format in use. A newer index format is in development which is substantially faster. (Robert Collins)
This release candidate has a few minor bug fixes, and some regression fixes for Windows.
BUG FIXES:
- bzr upgrade on remote branches accessed via bzr:// and bzr+ssh:// now works. (Andrew Bennetts)
- Change the get_format_description() strings for RepositoryFormatKnitPack5 et al to be single line messages. (Aaron Bentley)
- Fix for a regression on Win32 where we would try to call os.listdir() on a file and not catch the exception properly. (Windows raises a different exception.) This would manifest in places like bzr rm file or bzr switch. (Mark Hammond, John Arbash Meinel)
- Inventory.copy() was failing to set the revision property for the root entry. (Jelmer Vernooij)
- sftp transport: added missing FileExists case to _translate_io_exception (Christophe Troestler, #123475)
- The help for bzr ignored now suggests bzr ls --ignored for scripting use. (Robert Collins, #3834)
- The default annotate logic will now always assign the last-modified value of a line to one of the revisions that modified it, rather than a merge revision. This would happen when both sides claimed to have modified the line resulting in the same text. The choice is arbitrary but stable, so merges in different directions will get the same results. (John Arbash Meinel, #232188)
This release candidate for bzr 1.6 solidifies the new branch stacking feature. Bazaar now recommends that users upgrade all knit repositories, because later formats are much faster. However, we plan to continue read/write and upgrade support for knit repostories for the forseeable future. Several other bugs and performance issues were fixed.
CHANGES:
- Knit format repositories are deprecated and bzr will now emit warnings whenever it encounters one. Use bzr upgrade to upgrade knit repositories to pack format. (Andrew Bennetts)
IMPROVEMENTS:
- bzr check can now be told which elements at a location it should check. (Daniel Watkins)
- Commit now supports --exclude (or -x) to exclude some files from the commit. (Robert Collins, #3117)
- Fetching data between repositories that have the same model but no optimised fetcher will not reserialise all the revisions, increasing performance. (Robert Collins, John Arbash Meinel)
- Give a more specific error when target branch is not reachable. (James Westby)
- Implemented a custom walkdirs_utf8 implementation for win32. This uses a pyrex extension to get direct access to the FindFirstFileW style apis, rather than using listdir + lstat. Shows a very strong improvement in commands like status and diff which have to iterate the working tree. Anywhere from 2x-6x faster depending on the size of the tree (bigger trees, bigger benefit.) (John Arbash Meinel)
- New registry for log properties handles and the method in LongLogFormatter to display the custom properties returned by the registered handlers. (Guillermo Gonzalez, #162469)
BUG FIXES:
- Add more tests that stacking does not create deltas spanning physical repository boundaries. (Martin Pool, #252428)
- Better message about incompatible repositories. (Martin Pool, #206258)
- bzr branch --stacked ensures the destination branch format can support stacking, even if the origin does not. (Martin Pool)
- bzr export no longer exports .bzrrules. (Ian Clatworthy)
- bzr serve --directory=/ now correctly allows the whole filesystem to be accessed on Windows, not just the root of the drive that Python is running from. (Adrian Wilkins, #240910)
- Deleting directories by hand before running bzr rm will not cause subsequent errors in bzr st and bzr commit. (Robert Collins, #150438)
- Fix a test case that was failing if encoding wasn't UTF-8. (John Arbash Meinel, #247585)
- Fix "no buffer space available" error when branching with the new smart server protocol to or from Windows. (Andrew Bennetts, #246180)
- Fixed problem in branching from smart server. (#249256, Michael Hudson, Martin Pool)
- Handle a file turning in to a directory in TreeTransform. (James Westby, #248448)
API CHANGES:
- MutableTree.commit has an extra optional keywork parameter exclude that will be unconditionally supplied by the command line UI - plugins that add tree formats may need an update. (Robert Collins)
- The API minimum version for plugin compatibility has been raised to 1.6 - there are significant changes throughout the code base. (Robert Collins)
- The generic fetch code now uses three attributes on Repository objects to control fetch. The streams requested are controlled via : _fetch_order and _fetch_uses_deltas. Setting these appropriately allows different repository implementations to recieve data in their optimial form. If the _fetch_reconcile is set then a reconcile operation is triggered at the end of the fetch. (Robert Collins)
- The put_on_disk and get_tar_item methods in InventoryEntry were deprecated. (Ian Clatworthy)
- Repository.is_shared doesn't take a read lock. It didn't need one in the first place (nobody cached the value, and RemoteRepository wasn't taking one either). This saves a round trip when probing Pack repositories, as they read the pack-names file when locked. And during probe, locking the repo isn't very useful. (John Arbash Meinel)
INTERNALS:
- bzrlib.branchbuilder.BranchBuilder is now much more capable of putting together a real history without having to create a full WorkingTree. It is recommended that tests that are not directly testing the WorkingTree use BranchBuilder instead. See BranchBuilder.build_snapshot or TestCaseWithMemoryTree.make_branch_builder. (John Arbash Meinel)
- bzrlib.builtins.internal_tree_files broken into two giving a new helper safe_relpath_files - used by the new exclude parameter to commit. (Robert Collins)
- Make it easier to introduce new WorkingTree formats. (Ian Clatworthy)
- The code for exporting trees was refactored not to use the deprecated InventoryEntry methods. (Ian Clatworthy)
- RuleSearchers return () instead of [] now when there are no matches. (Ian Clatworthy)
This release adds a new 'stacked branches' feature allowing branches to share storage without being in the same repository or on the same machine. (See the user guide for more details.) It also adds a new hook, improved weaves, aliases for related locations, faster bzr+ssh push, and several bug fixes.
FEATURES:
- New pre_change_branch_tip hook that is called before the branch tip is moved, while the branch is write-locked. See the User Reference for signature details. (Andrew Bennetts)
- Rule-based preferences can now be defined for selected files in selected branches, allowing commands and plugins to provide custom behaviour for files matching defined patterns. See Rule-based preferences (part of Configuring Bazaar) in the User Guide and bzr help rules for more information. (Ian Clatworthy)
- Sites may suggest a branch to stack new branches on. (Aaron Bentley)
- Stacked branches are now supported. See bzr help branch and bzr help push. Branches must be in the development1 format to stack, though the stacked-on branch can be of any format. (Robert Collins)
IMPROVEMENTS:
- bzr export --format=tgz --root=NAME - to export a gzipped tarball to stdout; also tar and tbz2. (Martin Pool)
- bzr (re)merge --weave will now use a standard Weave algorithm, rather than the annotation-based merge it was using. It does so by building up a Weave of the important texts, without needing to build the full ancestry. (John Arbash Meinel, #238895)
- bzr send documents and better supports emacsclient (proper escaping of mail headers and handling of the MUA Mew). (Christophe Troestler)
- Remembered locations can be specified by aliases, e.g. :parent, :public, :submit. (Aaron Bentley)
- The smart protocol now has improved support for setting branches' revision info directly. This makes operations like push faster. The new request method name is Branch.set_last_revision_ex. (Andrew Bennetts)
BUG FIXES:
- Bazaar is now able to be a client to the web server of IIS 6 and 7. The broken implementations of RFC822 in Python and RFC2046 in IIS combined with boundary-line checking in Bazaar previously made this impossible. (NB, IIS 5 does not suffer from this problem). (Adrian Wilkins, #247585)
- bzr log --long with a ghost in your mainline now handles that ghost properly. (John Arbash Meinel, #243536)
- check handles the split-up .bzr layout correctly, so no longer requires a branch to be present. (Daniel Watkins, #64783)
- Clearer message about how to set the PYTHONPATH if bzrlib can't be loaded. (Martin Pool, #205230)
- Errors about missing libraries are now shown without a traceback, and with a suggestion to install the library. The full traceback is still in .bzr.log and can be shown with -Derror. (Martin Pool, #240161)
- Fetch from a stacked branch copies all required data. (Aaron Bentley, #248506)
- Handle urls such as ftp://user@host.com@www.host.com where the user name contains an @. (Neil Martinsen-Burrell, #228058)
- needs_read_lock and needs_write_lock now suppress an error during unlock if there was an error in the original function. This helps most when there is a failure with a smart server action, since often the connection closes and we cannot unlock. (Andrew Bennetts, John Arbash Meinel, #125784)
- Obsolete hidden command bzr fetch removed. (Martin Pool, #172870)
- Raise the correct exception when doing -rbefore:0 or -c0. (John Arbash Meinel, #239933)
- You can now compare file revisions in Windows diff programs from Cygwin Bazaar. (Matt McClure, #209281)
- revision_history now tolerates mainline ghosts for Branch format 6. (Aaron Bentley, #235055)
- Set locale from environment for third party libs. (Martin von Gagern, #128496)
DOCUMENTATION:
- Added Using stacked branches to the User Guide. (Ian Clatworthy)
- Updated developer documentation. (Martin Pool)
TESTING:
- -Dmemory will cause /proc/PID/status to be catted before bzr exits, allowing low-key analysis of peak memory use. (Robert Collins)
- TestCaseWithTransport.make_branch_and_tree tries harder to return a tree with a branch attribute of the right format. This was preventing some RemoteBranch tests from actually running with RemoteBranch instances. (Andrew Bennetts)
API CHANGES:
- Removed Repository.text_store, control_store, etc. Instead, there are new attributes texts, inventories, revisions, signatures, each of which is a VersionedFiles. See the Repository docstring for more details. (Robert Collins)
- Branch.pull now accepts an _override_hook_target optional parameter. If you have a subclass of Branch that overrides pull then you should add this parameter. (Andrew Bennetts)
- bzrlib.check.check() has been deprecated in favour of the more aptly-named bzrlib.check.check_branch(). (Daniel Watkins)
- Tree.print_file and Repository.print_file are deprecated. These methods are bad APIs because they write directly to sys.stdout. bzrlib does not use them internally, and there are no direct tests for them. (Alexander Belchenko)
INTERNALS:
- cat command no longer uses Tree.print_file() internally. (Alexander Belchenko)
- New class method BzrDir.open_containing_tree_branch_or_repository which eases the discovery of the tree, the branch and the repository containing a given location. (Daniel Watkins)
- New versionedfile.KeyMapper interface to abstract out the access to underlying .knit/.kndx etc files in repositories with partitioned storage. (Robert Collins)
- Obsolete developer-use command weave-join has been removed. (Robert Collins)
- RemoteToOtherFetcher and get_data_stream_for_search removed, to support new VersionedFiles layering. (Robert Collins)
This release contains further progress towards our 1.6 goals of shallow repositories, and contains a fix for some user-affecting bugs in the repository layer. Building working trees during checkout and branch is now faster.
BUG FIXES:
- Avoid KnitCorrupt error extracting inventories from some repositories. (The data is not corrupt; an internal check is detecting a problem reading from the repository.) (Martin Pool, Andrew Bennetts, Robert Collins, #234748)
- bzr status was breaking if you merged the same revision twice. (John Arbash Meinel, #235407)
- Fix infinite loop consuming 100% CPU when a connection is lost while reading a response body via the smart protocol v1 or v2. (Andrew Bennetts)
- Inserting a bundle which changes the contents of a file with no trailing end of line, causing a knit snapshot in a 'knits' repository will no longer cause KnitCorrupt. (Robert Collins)
- RemoteBranch.pull needs to return the self._real_branch's pull result. It was instead just returning None, which breaks bzr pull. (John Arbash Meinel, #238149)
- Sanitize branch nick before using it as an attachment filename in bzr send. (Lukáš Lalinský, #210218)
- Squash inv_entry.symlink_target to a plain string when generating DirState details. This prevents from getting a UnicodeError when you have symlinks and non-ascii filenames. (John Arbash Meinel, #135320)
IMPROVEMENTS:
- Added the 'alias' command to set/unset and display aliases. (Tim Penhey)
- added, modified, and unknowns behaviour made consistent (all three now quote paths where required). Added --null option to added and modified (for null-separated unknowns, use ls --unknown --null) (Adrian Wilkins)
- Faster branching (1.09x) and lightweight checkouts (1.06x) on large trees. (Ian Clatworthy, Aaron Bentley)
DOCUMENTATION:
- Added Bazaar Zen section to the User Guide. (Ian Clatworthy)
TESTING:
- Fix the test HTTPServer to be isolated from chdir calls made while it is running, allowing it to be used in blackbox tests. (Robert Collins)
API CHANGES:
- WorkingTree.set_parent_(ids/trees) will now filter out revisions which are in the ancestry of other revisions. So if you merge the same tree twice, or merge an ancestor of an existing merge, it will only record the newest. (If you merge a descendent, it will replace its ancestor). (John Arbash Meinel, #235407)
- RepositoryPolicy.__init__ now requires stack_on and stack_on_pwd, through the derived classes do not. (Aaron Bentley)
INTERNALS:
- bzrlib.bzrdir.BzrDir.sprout now accepts stacked to control creating stacked branches. (Robert Collins)
- Knit record serialisation is now stricter on what it will accept, to guard against potential internal bugs, or broken input. (Robert Collins)
Commands that work on the revision history such as push, pull, missing, uncommit and log are now substantially faster. This release adds a translation of some of the user documentation into Spanish. (Contributions of other translations would be very welcome.) Bazaar 1.6beta1 adds a new network protocol which is used by default and which allows for more efficient transfers and future extensions.
NOTES WHEN UPGRADING:
- There is a new version of the network protocol used for bzr://, bzr+ssh:// and bzr+http:// connections. This will allow more efficient requests and responses, and more graceful fallback when a server is too old to recognise a request from a more recent client. Bazaar 1.6 will interoperate with 0.16 and later versions, but servers should be upgraded when possible. Bazaar 1.6 no longer interoperates with 0.15 and earlier via these protocols. Use alternatives like SFTP or upgrade those servers. (Andrew Bennetts, #83935)
CHANGES:
- Deprecation warnings will not be suppressed when running bzr selftest so that developers can see if their code is using deprecated functions. (John Arbash Meinel)
FEATURES:
- Adding -Derror will now display a traceback when a plugin fails to load. (James Westby)
IMPROVEMENTS:
- bzr branch/push/pull -r XXX now have a helper function for finding the revno of the new revision (Graph.find_distance_to_null). This should make something like bzr branch -r -100 in a shared, no-trees repository much snappier. (John Arbash Meinel)
- bzr log --short -r X..Y no longer needs to access the full revision history. This makes it noticeably faster when logging the last few revisions. (John Arbash Meinel)
- bzr ls now accepts -V as an alias for --versioned. (Jerad Cramp, #165086)
- bzr missing uses the new Graph.find_unique_ancestors and Graph.find_differences to determine missing revisions without having to search the whole ancestry. (John Arbash Meinel, #174625)
- bzr uncommit now uses partial history access, rather than always extracting the full revision history for a branch. This makes it resolve the appropriate revisions much faster (in testing it drops uncommit from 1.5s => 0.4s). It also means bzr log --short is one step closer to not using full revision history. (John Arbash Meinel, #172649)
BUGFIXES:
- bzr merge --lca should handle when two revisions have no common ancestor other than NULL_REVISION. (John Arbash Meinel, #235715)
- bzr status was breaking if you merged the same revision twice. (John Arbash Meinel, #235407)
- bzr push with both --overwrite and -r NNN options no longer fails. (Andrew Bennetts, #234229)
- Correctly track the base URL of a smart medium when using bzr+http:// URLs, which was causing spurious "No repository present" errors with branches in shared repositories accessed over bzr+http. (Andrew Bennetts, #230550)
- Define _remote_is_at_least_1_2 on SmartClientMedium so that all implementations have the attribute. Fixes 'PyCurlTransport' object has no attribute '_remote_is_at_least_1_2' attribute errors. (Andrew Bennetts, #220806)
- Failure to delete an obsolete pack file should just give a warning message, not a fatal error. It may for example fail if the file is still in use by another process. (Martin Pool)
- Fix MemoryError during large fetches over HTTP by limiting the amount of data we try to read per recv call. The problem was observed with Windows and a proxy, but might affect other environments as well. (Eric Holmberg, #215426)
- Handle old merge directives correctly in Merger.from_mergeable. Stricter get_parent_map requirements exposed a latent bug here. (Aaron Bentley)
- Issue a warning and ignore passwords declared in authentication.conf when used for an ssh scheme (sftp or bzr+ssh). (Vincent Ladeuil, #203186)
- Make both http implementations raise appropriate exceptions on 403 Forbidden when POSTing smart requests. (Vincent Ladeuil, #230223)
- Properly title header names in http requests instead of capitalizing them. (Vincent Ladeuil, #229076)
- The "Unable to obtain lock" error message now also suggests using bzr break-lock to fix it. (Martin Albisetti, #139202)
- Treat an encoding of '' as ascii; this can happen when bzr is run under vim on Mac OS X. (Neil Martinsen-Burrell)
- VersionedFile.make_mpdiffs() was raising an exception that wasn't in scope. (Daniel Fischer #235687)
DOCUMENTATION:
- Added directory structure and started translation of docs in spanish. (Martin Albisetti, Lucio Albenga)
- Incorporate feedback from Jelmer Vernooij and Neil Martinsen-Burrell on the plugin and integration chapters of the User Guide. (Ian Clatworthy)
- More Bazaar developer documentation about packaging and release process, and about use of Python reprs. (Martin Pool, Martin Albisetti)
- Updated Tortise strategy document. (Mark Hammond)
TESTING:
- bzrlib.tests.adapt_tests was broken and unused - it has been fixed. (Robert Collins)
- Fix the test HTTPServer to be isolated from chdir calls made while it is running, allowing it to be used in blackbox tests. (Robert Collins)
- New helper function for splitting test suites split_suite_by_condition. (Robert Collins)
INTERNALS:
- Branch.missing_revisions has been deprecated. Similar functionality can be obtained using bzrlib.missing.find_unmerged. The api was fairly broken, and the function was unused, so we are getting rid of it. (John Arbash Meinel)
API CHANGES:
- Branch.abspath is deprecated; use the Tree or Transport instead. (Martin Pool)
- Branch.update_revisions now takes an optional Graph object. This can be used by update_revisions when it is checking ancestry, and allows callers to prefer request to go to a local branch. (John Arbash Meinel)
- Branch, Repository, Tree and BzrDir should expose a Transport as an attribute if they have one, rather than having it indirectly accessible as .control_files._transport. This doesn't add a requirement to support a Transport in cases where it was not needed before; it just simplifies the way it is reached. (Martin Pool)
- bzr missing --mine-only will return status code 0 if you have no new revisions, but the remote does. Similarly for --theirs-only. The new code only checks one side, so it doesn't know if the other side has changes. This seems more accurate with the request anyway. It also changes the output to print '[This|Other] branch is up to date.' rather than displaying nothing. (John Arbash Meinel)
- LockableFiles.put_utf8, put_bytes and controlfilename are now deprecated in favor of using Transport operations. (Martin Pool)
- Many methods on VersionedFile, Repository and in bzrlib.revision deprecated before bzrlib 1.5 have been removed. (Robert Collins)
- RevisionSpec.wants_revision_history can be set to False for a given RevisionSpec. This will disable the existing behavior of passing in the full revision history to self._match_on. Useful for specs that don't actually need access to the full history. (John Arbash Meinel)
- The constructors of SmartClientMedium and its subclasses now require a base parameter. SmartClientMedium implementations now also need to provide a remote_path_from_transport method. (Andrew Bennetts)
- The default permissions for creating new files and directories should now be obtained from BzrDir._get_file_mode() and _get_dir_mode(), rather than from LockableFiles. The _set_file_mode and _set_dir_mode variables on LockableFiles which were advertised as a way for plugins to control this are no longer consulted. (Martin Pool)
- VersionedFile.join is deprecated. This method required local instances of both versioned file objects and was thus hostile to being used for streaming from a smart server. The new get_record_stream and insert_record_stream are meant to efficiently replace this method. (Robert Collins)
- WorkingTree.set_parent_(ids/trees) will now filter out revisions which are in the ancestry of other revisions. So if you merge the same tree twice, or merge an ancestor of an existing merge, it will only record the newest. (If you merge a descendent, it will replace its ancestor). (John Arbash Meinel, #235407)
- WorkingTreeFormat2.stub_initialize_remote is now private. (Martin Pool)
This release of Bazaar includes several updates to the documentation, and fixes to prepare for making rich root support the default format. Many bugs have been squashed, including fixes to log, bzr+ssh inter-operation with older servers.
CHANGES:
- Suppress deprecation warnings when bzrlib is a 'final' release. This way users of packaged software won't be bothered with DeprecationWarnings, but developers and testers will still see them. (John Arbash Meinel)
DOCUMENTATION:
- Incorporate feedback from Jelmer Vernooij and Neil Martinsen-Burrell on the plugin and integration chapters of the User Guide. (Ian Clatworthy)
NOTES WHEN UPGRADING:
CHANGES:
- Broader support of GNU Emacs mail clients. Set mail_client=emacsclient in your bazaar.conf and send will pop the bundle in a mail buffer according to the value of mail-user-agent variable. (Xavier Maillard)
FEATURES:
IMPROVEMENTS:
- Diff now handles revision specs like "branch:" and "submit:" more efficiently. (Aaron Bentley, #202928)
- More friendly error given when attempt to start the smart server on an address already in use. (Andrea Corbellini, #200575)
- Pull completes much faster when there is nothing to pull. (Aaron Bentley)
BUGFIXES:
- Authentication.conf can define sections without password. (Vincent Ladeuil, #199440)
- Avoid muttering every time a child update does not cause a progress bar update. (John Arbash Meinel, #213771)
- Branch.reconcile() is now implemented. This allows bzr reconcile to fix when a Branch has a non-canonical mainline history. bzr check also detects this condition. (John Arbash Meinel, #177855)
- bzr log -r ..X bzr:// was failing, because it was getting a request for revision_id=None which was not a string. (John Arbash Meinel, #211661)
- bzr commit now works with Microsoft's FTP service. (Andreas Deininger)
- Catch definitions outside sections in authentication.conf. (Vincent Ladeuil, #217650)
- Conversion from non-rich-root to rich-root(-pack) updates inventory sha1s, even when bundles are used. (Aaron Bentley, #181391)
- Conversion from non-rich-root to rich-root(-pack) works correctly even though search keys are not topologically sorted. (Aaron Bentley)
- Conversion from non-rich-root to rich-root(-pack) works even when a parent revision has a different root id. (Aaron Bentley, #177874)
- Disable strace testing until strace is fixed (see bug #103133) and emit a warning when selftest ends to remind us of leaking tests. (Vincent Ladeuil, #226769)
- Fetching all revisions from a repository does not cause pack collisions. (Robert Collins, Aaron Bentley, #212908)
- Fix error about "attempt to add line-delta in non-delta knit". (Andrew Bennetts, #217701)
- Pushing a branch in "dirstate" format (Branch5) over bzr+ssh would break if the remote server was < version 1.2. This was due to a bug in the RemoteRepository.get_parent_map() fallback code. (John Arbash Meinel, #214894)
- Remove leftover code in bzr_branch that inappropriately creates a branch-name file in the branch control directory. (Martin Pool)
- Set SO_REUSEADDR on server sockets of bzr serve to avoid problems rebinding the socket when starting the server a second time. (John Arbash Meinel, Martin Pool, #164288)
- Severe performance degradation in fetching from knit repositories to knits and packs due to parsing the entire revisions.kndx on every graph walk iteration fixed by using the Repository.get_graph API. There was another regression in knit => knit fetching which re-read the index for every revision each side had in common. (Robert Collins, John Arbash Meinel)
- When logging the changes to a particular file, there was a bug if there were ghosts in the revision ancestry. (John Arbash Meinel, #209948)
- xs4all's ftp server returns a temporary error when trying to list an empty directory, rather than returning an empty list. Adding a workaround so that we don't get spurious failures. (John Arbash Meinel, #215522)
DOCUMENTATION:
- Expanded the User Guide to include new chapters on popular plugins and integrating Bazaar into your environment. The Best practices chapter was renamed to Miscellaneous topics as suggested by community feedback as well. (Ian Clatworthy)
- Document outlining strategies for TortoiseBzr. (Mark Hammond)
- Improved the documentation on hooks. (Ian Clatworthy)
- Update authentication docs regarding ssh agents. (Vincent Ladeuil, #183705)
TESTING:
- Add thread_name_suffix parameter to SmartTCPServer_for_testing, to make it easy to identify which test spawned a thread with an unhandled exception. (Andrew Bennetts)
- New --debugflag/-E option to bzr selftest for setting options for debugging tests, these are complementary to the the -D options. The -Dselftest_debug global option has been replaced by the -E=allow_debug option for selftest. (Andrew Bennetts)
- Parameterised test ids are preserved correctly to aid diagnosis of test failures. (Robert Collins, Andrew Bennetts)
- selftest now accepts --starting-with <id> to load only the tests whose id starts with the one specified. This greatly speeds up running the test suite on a limited set of tests and can be used to run the tests for a single module, a single class or even a single test. (Vincent Ladeuil)
- The test suite modules have been modified to define load_tests() instead of test_suite(). That speeds up selective loading (via --load-list) significantly and provides many examples on how to migrate (grep for load_tests). (Vincent Ladeuil)
INTERNALS:
- Hooks.install_hook is now deprecated in favour of Hooks.install_named_hook which adds a required name parameter, to avoid having to call Hooks.name_hook. (Daniel Watkins)
- Implement xml8 serializer. (Aaron Bentley)
- New form @deprecated_method(deprecated_in(1, 5, 0)) for making deprecation wrappers. (Martin Pool)
- Repository.revision_parents is now deprecated in favour of Repository.get_parent_map([revid])[revid]. (Jelmer Vernooij)
- The Python assert statement is no longer used in Bazaar source, and a test checks this. (Martin Pool)
API CHANGES:
- bzrlib.status.show_pending_merges requires the repository to be locked by the caller. Callers should have been doing it anyway, but it will now raise an exception if they do not. (John Arbash Meinel)
- Repository.get_data_stream, Repository.get_data_stream_for_search(), Repository.get_deltas_for_revsions(), Repository.revision_trees(), Repository.item_keys_introduced_by() no longer take read locks. (Aaron Bentley)
- LockableFiles.get_utf8 and .get are deprecated, as a start towards removing LockableFiles and .control_files entirely. (Martin Pool)
- Methods deprecated prior to 1.1 have been removed. (Martin Pool)
This release of Bazaar includes handy improvements to the speed of log and status, new options for several commands, improved documentation, and better hooks, including initial code for server-side hooks. A number of bugs have been fixed, particularly in interoperability between different formats or different releases of Bazaar over there network. There's been substantial internal work in both the repository and network code to enable new features and faster performance.
BUG FIXES:
- Pushing a branch in "dirstate" format (Branch5) over bzr+ssh would break if the remote server was < version 1.2. This was due to a bug in the RemoteRepository.get_parent_map() fallback code. (John Arbash Meinel, Andrew Bennetts, #214894)
BUG FIXES:
- bzr log -r ..X bzr:// was failing, because it was getting a request for revision_id=None which was not a string. (John Arbash Meinel, #211661)
- Fixed a bug in handling ghost revisions when logging changes in a particular file. (John Arbash Meinel, #209948)
- Fix error about "attempt to add line-delta in non-delta knit". (Andrew Bennetts, #205156)
- Fixed performance degradation in fetching from knit repositories to knits and packs due to parsing the entire revisions.kndx on every graph walk iteration fixed by using the Repository.get_graph API. There was another regression in knit => knit fetching which re-read the index for every revision each side had in common. (Robert Collins, John Arbash Meinel)
CHANGES:
- bzr main script cannot be imported (Benjamin Peterson)
- On Linux bzr additionally looks for plugins in arch-independent site directory. (Toshio Kuratomi)
- The set_rh branch hook is now deprecated. Please migrate any plugins using this hook to use an alternative, e.g. post_change_branch_tip. (Ian Clatworthy)
- When a plugin cannot be loaded as the file path is not a valid python module name bzr will now strip a bzr_ prefix from the front of the suggested name, as many plugins (e.g. bzr-svn) want to be installed without this prefix. It is a common mistake to have a folder named "bzr-svn" for that plugin, especially as this is what bzr branch lp:bzr-svn will give you. (James Westby, Andrew Cowie)
- UniqueIntegerBugTracker now appends bug-ids instead of joining them to the base URL. Plugins that register bug trackers may need a trailing / added to the base URL if one is not already there. (James Wesby, Andrew Cowie)
FEATURES:
- Added start_commit hook for mutable trees. (Jelmer Vernooij, #186422)
- status now accepts --no-pending to show the status without listing pending merges, which speeds up the command a lot on large histories. (James Westby, #202830)
- New post_change_branch_tip hook that is called after the branch tip is moved but while the branch is still write-locked. See the User Reference for signature details. (Ian Clatworthy, James Henstridge)
- Reconfigure can convert a branch to be standalone or to use a shared repository. (Aaron Bentley)
IMPROVEMENTS:
- The smart protocol now has support for setting branches' revision info directly. This should make operations like push slightly faster, and is a step towards server-side hooks. The new request method name is Branch.set_last_revision_info. (Andrew Bennetts)
- bzr commit --fixes now recognises "gnome" as a tag by default. (James Westby, Andrew Cowie)
- bzr switch will attempt to find branches to switch to relative to the current branch. E.g. bzr switch branchname will look for current_branch/../branchname. (Robert Collins, Jelmer Vernooij, Wouter van Heyst)
- Diff is now more specific about execute-bit changes it describes (Chad Miller)
- Fetching data over HTTP is a bit faster when urllib is used. This is done by forcing it to recv 64k at a time when reading lines in HTTP headers, rather than just 1 byte at a time. (Andrew Bennetts)
- Log --short and --line are much faster when -r is not specified. (Aaron Bentley)
- Merge is faster. We no longer check a file's existence unnecessarily when merging the execute bit. (Aaron Bentley)
- bzr status on an explicit list of files no longer shows pending merges, making it much faster on large trees. (John Arbash Meinel)
- The launchpad directory service now warns the user if they have not set their launchpad login and are trying to resolve a URL using it, just in case they want to do a write operation with it. (James Westby)
- The smart protocol client is slightly faster, because it now only queries the server for the protocol version once per connection. Also, the HTTP transport will now automatically probe for and use a smart server if one is present. You can use the new nosmart+ transport decorator to get the old behaviour. (Andrew Bennetts)
- The version command takes a --short option to print just the version number, for easier use in scripts. (Martin Pool)
- Various operations with revision specs and commands that calculate revnos and revision ids are faster. (John A. Meinel, Aaron Bentley)
BUGFIXES:
- Add root_client_path parameter to SmartWSGIApp and SmartServerRequest. This makes it possible to publish filesystem locations that don't exactly match URL paths. SmartServerRequest subclasses should use the new translate_client_path and transport_from_client_path methods when dealing with paths received from a client to take this into account. (Andrew Bennetts, #124089)
- bzr mv a b can be now used also to rename previously renamed directories, not only files. (Lukáš Lalinský, #107967)
- bzr uncommit --local can now remove revisions from the local branch to be symmetric with bzr commit --local. (John Arbash Meinel, #93412)
- Don't ask for a password if there is no real terminal. (Alexander Belchenko, #69851)
- Fix a bug causing a ValueError crash in parse_line_delta_iter when fetching revisions from a knit to pack repository or vice versa using bzr:// (including over http or ssh). (#208418, Andrew Bennetts, Martin Pool, Robert Collins)
- Fixed _get_line in bzrlib.smart.medium, which was buggy. Also fixed _get_bytes in the same module to use the push back buffer. These bugs had no known impact in normal use, but were problematic for developers working on the code, and were likely to cause real bugs sooner or later. (Andrew Bennetts)
- Implement handling of basename parameter for DefaultMail. (James Westby)
- Incompatibility with Paramiko versions newer than 1.7.2 was fixed. (Andrew Bennetts, #213425)
- Launchpad locations (lp: URLs) can be pulled. (Aaron Bentley, #181945)
- Merges that add files to deleted root directories complete. They do create conflicts. (Aaron Bentley, #210092)
- vsftp's return 550 RNFR command failed. supported. (Marcus Trautwig, #129786)
DOCUMENTATION:
- Improved documentation on send/merge relationship. (Peter Schuller)
- Minor fixes to the User Guide. (Matthew Fuller)
- Reduced the evangelism in the User Guide. (Ian Clatworthy)
- Added Integrating with Bazaar document for developers (Martin Albisetti)
API BREAKS:
- Attempting to pull data from a ghost aware repository (e.g. knits) into a non-ghost aware repository such as weaves will now fail if there are ghosts. (Robert Collins)
- KnitVersionedFile no longer accepts an access_mode parameter, and now requires the index and access_method parameters to be supplied. A compatible shim has been kept in the new function knit.make_file_knit. (Robert Collins)
- Log formatters must now provide log_revision instead of show and show_merge_revno methods. The latter had been deprecated since the 0.17 release. (James Westby)
- LoopbackSFTP is now called SocketAsChannelAdapter. (Andrew Bennetts)
- osutils.backup_file is removed. (Alexander Belchenko)
- Repository.get_revision_graph is deprecated, with no replacement method. The method was size(history) and not desirable. (Robert Collins)
- revision.revision_graph is deprecated, with no replacement function. The function was size(history) and not desirable. (Robert Collins)
- Transport.get_shared_medium is deprecated. Use Transport.get_smart_medium instead. (Andrew Bennetts)
- VersionedFile factories now accept a get_scope parameter rather than using a call to transaction_finished, allowing the removal of the fixed list of versioned files per repository. (Robert Collins)
- VersionedFile.annotate_iter is deprecated. While in principle this allowed lower memory use, all users of annotations wanted full file annotations, and there is no storage format suitable for incremental line-by-line annotation. (Robert Collins)
- VersionedFile.clone_text is deprecated. This performance optimisation is no longer used - reading the content of a file that is undergoing a file level merge to identical state on two branches is rare enough, and not expensive enough to special case. (Robert Collins)
- VersionedFile.clear_cache and enable_cache are deprecated. These methods added significant complexity to the VersionedFile implementation, but were only used for optimising fetches from knits - which can be done from outside the knit layer, or via a caching decorator. As knits are not the default format, the complexity is no longer worth paying. (Robert Collins)
- VersionedFile.create_empty is removed. This method presupposed a sensible mapping to a transport for individual files, but pack backed versioned files have no such mapping. (Robert Collins)
- VersionedFile.get_graph is deprecated, with no replacement method. The method was size(history) and not desirable. (Robert Collins)
- VersionedFile.get_graph_with_ghosts is deprecated, with no replacement method. The method was size(history) and not desirable. (Robert Collins)
- VersionedFile.get_parents is deprecated, please use VersionedFile.get_parent_map. (Robert Collins)
- VersionedFile.get_sha1 is deprecated, please use VersionedFile.get_sha1s. (Robert Collins)
- VersionedFile.has_ghost is now deprecated, as it is both expensive and unused outside of a single test. (Robert Collins)
- VersionedFile.iter_parents is now deprecated in favour of get_parent_map which can be used to instantiate a Graph on a VersionedFile. (Robert Collins)
- VersionedFileStore no longer uses the transaction parameter given to most methods; amongst other things this means that the get_weave_or_empty method no longer guarantees errors on a missing weave in a readonly transaction, and no longer caches versioned file instances which reduces memory pressure (but requires more careful management by callers to preserve performance). (Robert Collins)
TESTING:
New -Dselftest_debug flag disables clearing of the debug flags during tests. This is useful if you want to use e.g. -Dhpss to help debug a failing test. Be aware that using this feature is likely to cause spurious test failures if used with the full suite. (Andrew Bennetts)
selftest --load-list now uses a new more agressive test loader that will avoid loading unneeded modules and building their tests. Plugins can use this new loader by defining a load_tests function instead of a test_suite function. (a forthcoming patch will provide many examples on how to implement this). (Vincent Ladeuil)
selftest --load-list now does some sanity checks regarding duplicate test IDs and tests present in the list but not found in the actual test suite. (Vincent Ladeuil)
Slightly more concise format for the selftest progress bar, so there's more space to show the test name. (Martin Pool)
[2500/10884, 1fail, 3miss in 1m29s] test_revisionnamespaces.TestRevThe test suite takes much less memory to run, and is a bit faster. This is done by clearing most attributes of TestCases after running them, if they succeeded. (Andrew Bennetts)
INTERNALS:
- Added _build_client_protocol to _SmartClient. (Andrew Bennetts)
- Added basic infrastructure for automatic plugin suggestion. (Martin Albisetti)
- If a LockableFiles object is not explicitly unlocked (for example because of a missing try/finally block, it will give a warning but not automatically unlock itself. (Previously they did.) This sometimes caused knock-on errors if for example the network connection had already failed, and should not be relied upon by code. (Martin Pool, #109520)
- make dist target to build a release tarball, and also check-dist-tarball and dist-upload-escudero. (Martin Pool)
- The read_response_tuple method of SmartClientRequestProtocol* classes will now raise UnknownSmartMethod when appropriate, so that callers don't need to try distinguish unknown request errors from other errors. (Andrew Bennetts)
- set_make_working_trees is now implemented provided on all repository implementations (Aaron Bentley)
- VersionedFile now has a new method get_parent_map which, like Graph.get_parent_map returns a dict of key:parents. (Robert Collins)
No changes from 1.3.1rc1.
BUG FIXES:
- Fix a bug causing a ValueError crash in parse_line_delta_iter when fetching revisions from a knit to pack repository or vice versa using bzr:// (including over http or ssh). (#208418, Andrew Bennetts, Martin Pool, Robert Collins)
Bazaar has become part of the GNU project <http://www.gnu.org>
Many operations that act on history, including log and annotate are now substantially faster. Several bugs have been fixed and several new options and features have been added.
TESTING:
- Avoid spurious failure of TestVersion.test_version matching directory names. (#202778, Martin Pool)
NOTES WHEN UPGRADING:
- The backup directory created by upgrade is now called backup.bzr, not .bzr.backup. (Martin Albisetti)
CHANGES:
- A new repository format 'development' has been added. This format will represent the latest 'in-progress' format that the bzr developers are interested in getting early-adopter testing and feedback on. doc/developers/development-repo.txt has detailed information. (Robert Collins)
- BZR_LOG environment variable controls location of .bzr.log trace file. User can suppress writing messages to .bzr.log by using '/dev/null' filename (on Linux) or 'NUL' (on Windows). If BZR_LOG variable is not defined but BZR_HOME is defined then default location for .bzr.log trace file is $BZR_HOME/.bzr.log. (Alexander Belchenko)
- launchpad builtin plugin now shipped as separate part in standalone bzr.exe, installed to C:\Program Files\Bazaar\plugins directory, and standalone installer allows user to skip installation of this plugin. (Alexander Belchenko)
- Restore auto-detection of plink.exe on Windows. (Dmitry Vasiliev)
- Version number is now shown as "1.2" or "1.2pr2", without zeroed or missing final fields. (Martin Pool)
FEATURES:
- branch and checkout can hard-link working tree files, which is faster and saves space. (Aaron Bentley)
- bzr send will now also look at the child_submit_to setting in the submit branch to determine the email address to send to. (Jelmer Vernooij)
IMPROVEMENTS:
- BzrBranch._lefthand_history is faster on pack repos. (Aaron Bentley)
- Branch6.generate_revision_history is faster. (Aaron Bentley)
- Directory services can now be registered, allowing special URLs to be dereferenced into real URLs. This is a generalization and cleanup of the lp: transport lookup. (Aaron Bentley)
- Merge directives that are automatically attached to emails have nicer filenames, based on branch-nick + revno. (Aaron Bentley)
- push has a --revision option, to specify what revision to push up to. (Daniel Watkins)
- Significantly reducing execution time and network traffic for trivial case of running bzr missing command for two identical branches. (Alexander Belchenko)
- Speed up operations that look at the revision graph (such as 'bzr log'). KnitPackRepositor.get_revision_graph uses Graph.iter_ancestry to extract the revision history. This allows filtering ghosts while stepping instead of needing to peek ahead. (John Arbash Meinel)
- The hooks command lists installed hooks, to assist in debugging. (Daniel Watkins)
- Updates to how annotate work. Should see a measurable improvement in performance and memory consumption for file with a lot of merges. Also, correctly handle when a line is introduced by both parents (it should be attributed to the first merge which notices this, and not to all subsequent merges.) (John Arbash Meinel)
BUGFIXES:
- Autopacking no longer holds the full set of inventory lines in memory while copying. For large repositories, this can amount to hundreds of MB of ram consumption. (Ian Clatworthy, John Arbash Meinel)
- Cherrypicking when using --format=merge3 now explictly excludes BASE lines. (John Arbash Meinel, #151731)
- Disable plink's interactive prompt for password. (#107593, Dmitry Vasiliev)
- Encode command line arguments from unicode to user_encoding before invoking external mail client in bzr send command. (#139318, Alexander Belchenko)
- Fixed problem connecting to bzr+https:// servers. (#198793, John Ferlito)
- Improved error reporting in the Launchpad plugin. (Daniel Watkins, #196618)
- Include quick-start-summary.svg file to python-based installer(s) for Windows. (#192924, Alexander Belchenko)
- lca merge now respects specified files. (Aaron Bentley)
- Make version-info --custom imply --all. (#195560, James Westby)
- merge --preview now works for merges that add or modify symlinks (James Henstridge)
- Redirecting the output from bzr merge (when the remembered location is used) now works. (John Arbash Meinel)
- setup.py script explicitly checks for Python version. (Jari Aalto, Alexander Belchenko, #200569)
- UnknownFormatErrors no longer refer to branches regardless of kind of unknown format. (Daniel Watkins, #173980)
- Upgrade bundled ConfigObj to version 4.5.2, which properly quotes # signs, among other small improvements. (Matt Nordhoff, #86838)
- Use correct indices when emitting LCA conflicts. This fixes IndexError errors. (Aaron Bentley, #196780)
DOCUMENTATION:
- Explained how to use version-info --custom in the User Guide. (Neil Martinsen-Burrell)
API BREAKS:
- Support for loading plugins from zip files and bzrlib.plugin.load_from_zip() function are deprecated. (Alexander Belchenko)
TESTING:
- Added missing blackbox tests for modified (Adrian Wilkins)
- The branch interface tests were invalid for branches using rich-root repositories because the empty string is not a valid file-id. (Robert Collins)
INTERNALS:
- Graph.iter_ancestry returns the ancestry of revision ids. Similar to Repository.get_revision_graph() except it includes ghosts and you can stop part-way through. (John Arbash Meinel)
- New module tools/package_mf.py provide custom module finder for python packages (improves standard python library's modulefinder.py) used by setup.py script while building standalone bzr.exe. (Alexander Belchenko)
- New remote method RemoteBzrDir.find_repositoryV2 adding support for detecting external lookup support on remote repositories. This method is now attempted first when lookup up repositories, leading to an extra round trip on older bzr smart servers. (Robert Collins)
- Repository formats have a new supported-feature attribute supports_external_lookups used to indicate repositories which support falling back to other repositories when they have partial data. (Robert Collins)
- Repository.get_revision_graph_with_ghosts and bzrlib.revision.(common_ancestor,MultipleRevisionSources,common_graph) have been deprecated. (John Arbash Meinel)
- Tree.iter_changes is now a public API, replacing the work-in-progress Tree._iter_changes. The api is now considered stable and ready for external users. (Aaron Bentley)
- The bzrdir format registry now accepts an alias keyword to register_metadir, used to indicate that a format name is an alias for some other format and thus should not be reported when describing the format. (Robert Collins)
BUG FIXES:
- Fix failing test in Launchpad plugin. (Martin Pool)
NOTES WHEN UPGRADING:
- Fetching via the smart protocol may need to reconnect once during a fetch if the remote server is running Bazaar 1.1 or earlier, because the client attempts to use more efficient requests that confuse older servers. You may be required to re-enter a password or passphrase when this happens. This won't happen if the server is upgraded to Bazaar 1.2. (Andrew Bennetts)
CHANGES:
- Fetching via bzr+ssh will no longer fill ghosts by default (this is consistent with pack-0.92 fetching over SFTP). (Robert Collins)
- Formatting of bzr plugins output is changed to be more human- friendly. Full path of plugins locations will be shown only with --verbose command-line option. (Alexander Belchenko)
- merge now prefers to use the submit branch, but will fall back to parent branch. For many users, this has no effect. But some users who pull and merge on the same branch will notice a change. This change makes it easier to work on a branch on two different machines, pulling between the machines, while merging from the upstream. merge --remember can now be used to set the submit_branch. (Aaron Bentley)
FEATURES:
- merge --preview produces a diff of the changes merge would make, but does not actually perform the merge. (Aaron Bentley)
- New smart method Repository.get_parent_map for getting revision parent data. This returns additional parent information topologically adjacent to the requested data to reduce round trip latency impacts. (Robert Collins)
- New smart method, Repository.stream_revisions_chunked, for fetching revision data that streams revision data via a chunked encoding. This avoids buffering large amounts of revision data on the server and on the client, and sends less data to the server to request the revisions. (Andrew Bennetts, Robert Collins, #178353)
- The launchpad plugin now handles lp urls of the form lp://staging/, lp://demo/, lp://dev/ to use the appropriate launchpad instance to do the resolution of the branch identities. This is primarily of use to Launchpad developers, but can also be used by other users who want to try out Launchpad as a branch location without messing up their public Launchpad account. Branches that are pushed to the staging environment have an expected lifetime of one day. (Tim Penhey)
IMPROVEMENTS:
- Creating a new branch no longer tries to read the entire revision-history unnecessarily over smart server operations. (Robert Collins)
- Fetching between different repository formats with compatible models now takes advantage of the smart method to stream revisions. (Andrew Bennetts)
- The --coverage option is now global, rather specific to bzr selftest. (Andrew Bennetts)
- The register-branch command will now use the public url of the branch containing the current directory, if one has been set and no explicit branch is provided. (Robert Collins)
- Tweak the reannotate code path to optimize the 2-parent case. Speeds up bzr annotate with a pack repository by approx 3:2. (John Arbash Meinel)
BUGFIXES:
- Calculate remote path relative to the shared medium in _SmartClient. This is related to the problem in bug #124089. (Andrew Bennetts)
- Cleanly handle connection errors in smart protocol version two, the same way as they are handled by version one. (Andrew Bennetts)
- Clearer error when version-info --custom is used without --template (Lukáš Lalinský)
- Don't raise UnavailableFeature during test setup when medusa is not available or tearDown is never called leading to nasty side effects. (#137823, Vincent Ladeuil)
- If a plugin's test suite cannot be loaded, for example because of a syntax error in the tests, then selftest fails, rather than just printing a warning. (Martin Pool, #189771)
- List possible values for BZR_SSH environment variable in env-variables help topic. (Alexander Belchenko, #181842)
- New methods push_log_file and pop_log_file to intercept messages: popping the log redirection now precisely restores the previous state, which makes it easier to use bzr log output from other programs. TestCaseInTempDir no longer depends on a log redirection being established by the test framework, which lets bzr tests cleanly run from a normal unittest runner. (#124153, #124849, Martin Pool, Jonathan Lange)
- pull --quiet is now more quiet, in particular a message is no longer printed when the remembered pull location is used. (James Westby, #185907)
- reconfigure can safely be interrupted while fetching. (Aaron Bentley, #179316)
- reconfigure preserves tags when converting to and from lightweight checkouts. (Aaron Bentley, #182040)
- Stop polluting /tmp when running selftest. (Vincent Ladeuil, #123623)
- Switch from NFKC => NFC for normalization checks. NFC allows a few more characters which should be considered valid. (John Arbash Meinel, #185458)
- The launchpad plugin now uses the edge xmlrpc server to avoid interacting badly with a bug on the launchpad side. (Robert Collins)
- Unknown hostnames when connecting to a bzr:// URL no longer cause tracebacks. (Andrew Bennetts, #182849)
API BREAKS:
- Classes implementing Merge types like Merge3Merger must now accept (and honour) a do_merge flag in their constructor. (Aaron Bentley)
- Repository.add_inventory and add_revision now require the caller to previously take a write lock (and start a write group.) (Martin Pool)
TESTING:
- selftest now accepts --load-list <file> to load a test id list. This speeds up running the test suite on a limited set of tests. (Vincent Ladeuil)
INTERNALS:
- Add a new method get_result to graph search objects. The resulting SearchResult can be used to recreate the search later, which will be useful in reducing network traffic. (Robert Collins)
- Use convenience function to check whether two repository handles are referring to the same repository in Repository.get_graph. (Jelmer Vernooij, #187162)
- Fetching now passes the find_ghosts flag through to the InterRepository.missing_revision_ids call consistently for all repository types. This will enable faster missing revision discovery with bzr+ssh. (Robert Collins)
- Fix error handling in Repository.insert_data_stream. (Lukas Lalinsky)
- InterRepository.missing_revision_ids is now deprecated in favour of InterRepository.search_missing_revision_ids which returns a bzrlib.graph.SearchResult suitable for making requests from the smart server. (Robert Collins)
- New error NoPublicBranch for commands that need a public branch to operate. (Robert Collins)
- New method iter_inventories on Repository for access to many inventories. This is primarily used by the revision_trees method, as direct access to inventories is discouraged. (Robert Collins)
- New method next_with_ghosts on the Graph breadth-first-search objects which will split out ghosts and present parents into two separate sets, useful for code which needs to be aware of ghosts (e.g. fetching data cares about ghosts during revision selection). (Robert Collins)
- Record a timestamp against each mutter to the trace file, relative to the first import of bzrlib. (Andrew Bennetts)
- Repository.get_data_stream is now deprecated in favour of Repository.get_data_stream_for_search which allows less network traffic when requesting data streams over a smart server. (Robert Collins)
- RemoteBzrDir._get_tree_branch no longer triggers _ensure_real, removing one round trip on many network operations. (Robert Collins)
- RemoteTransport's recommended_page_size method now returns 64k, like SFTPTransport and HttpTransportBase. (Andrew Bennetts)
- Repository has a new method has_revisions which signals the presence of many revisions by returning a set of the revisions listed which are present. This can be done by index queries without reading data for parent revision names etc. (Robert Collins)
(no changes from 1.1rc1)
CHANGES:
- Dotted revision numbers have been revised. Instead of growing longer with nested branches the branch number just increases. (eg instead of 1.1.1.1.1 we now report 1.2.1.) This helps scale long lived branches which have many feature branches merged between them. (John Arbash Meinel)
- The syntax bzr diff branch1 branch2 is no longer supported. Use bzr diff branch1 --new branch2 instead. This change has been made to remove the ambiguity where branch2 is in fact a specific file to diff within branch1.
FEATURES:
- New option to use custom template-based formats in bzr version-info. (Lukáš Lalinský)
- diff '--using' allows an external diff tool to be used for files. (Aaron Bentley)
- New "lca" merge-type for fast everyday merging that also supports criss-cross merges. (Aaron Bentley)
IMPROVEMENTS:
- annotate now doesn't require a working tree. (Lukáš Lalinský, #90049)
- branch and checkout can now use files from a working tree to to speed up the process. For checkout, this requires the new --files-from flag. (Aaron Bentley)
- bzr diff now sorts files in alphabetical order. (Aaron Bentley)
- bzr diff now works on branches without working trees. Tree-less branches can also be compared to each other and to working trees using the new diff options --old and --new. Diffing between branches, with or without trees, now supports specific file filtering as well. (Ian Clatworthy, #6700)
- bzr pack now orders revision texts in topological order, with newest at the start of the file, promoting linear reads for bzr log and the like. This partially fixes #154129. (Robert Collins)
- Merge directives now fetch prerequisites from the target branch if needed. (Aaron Bentley)
- pycurl now handles digest authentication. (Vincent Ladeuil)
- reconfigure can now convert from repositories. (Aaron Bentley)
- -l is now a short form for --limit in log. (Matt Nordhoff)
- merge now warns when merge directives cause cherrypicks. (Aaron Bentley)
- split now supported, to enable splitting large trees into smaller pieces. (Aaron Bentley)
BUGFIXES:
- Avoid AttributeError when unlocking a pack repository when an error occurs. (Martin Pool, #180208)
- Better handle short reads when processing multiple range requests. (Vincent Ladeuil, #179368)
- build_tree acceleration uses the correct path when a file has been moved. (Aaron Bentley)
- commit now succeeds when a checkout and its master branch share a repository. (Aaron Bentley, #177592)
- Fixed error reporting of unsupported timezone format in log --timezone. (Lukáš Lalinský, #178722)
- Fixed Unicode encoding error in ignored when the output is redirected to a pipe. (Lukáš Lalinský)
- Fix traceback when sending large response bodies over the smart protocol on Windows. (Andrew Bennetts, #115781)
- Fix urlutils.relative_url for the case of two file:/// URLs pointed to different logical drives on Windows. (Alexander Belchenko, #90847)
- HTTP test servers are now compatible with the http protocol version 1.1. (Vincent Ladeuil, #175524)
- _KnitParentsProvider.get_parent_map now handles requests for ghosts correctly, instead of erroring or attributing incorrect parents to ghosts. (Aaron Bentley)
- merge --weave --uncommitted now works. (Aaron Bentley)
- pycurl authentication handling was broken and incomplete. Fix handling of user:pass embedded in the urls. (Vincent Ladeuil, #177643)
- Files inside non-directories are now handled like other conflict types. (Aaron Bentley, #177390)
- reconfigure is able to convert trees into lightweight checkouts. (Aaron Bentley)
- Reduce lockdir timeout to 0 when running bzr serve. (Andrew Bennetts, #148087)
- Test that the old version_info_format functions still work, even though they are deprecated. (John Arbash Meinel, ShenMaq, #177872)
- Transform failures no longer cause ImmortalLimbo errors (Aaron Bentley, #137681)
- uncommit works even when the commit messages of revisions to be removed use characters not supported in the terminal encoding. (Aaron Bentley)
- When dumb http servers return whole files instead of the requested ranges, read the remaining bytes by chunks to avoid overflowing network buffers. (Vincent Ladeuil, #175886)
DOCUMENTATION:
- Minor tweaks made to the bug tracker integration documentation. (Ian Clatworthy)
- Reference material has now be moved out of the User Guide and added to the User Reference. The User Reference has gained 4 sections as a result: Authenication Settings, Configuration Settings, Conflicts and Hooks. All help topics are now dumped into text format in the doc/en/user-reference directory for those who like browsing that information in their editor. (Ian Clatworthy)
- Using Bazaar with Launchpad tutorial added. (Ian Clatworthy)
INTERNALS:
- find_* methods available for BzrDirs, Branches and WorkingTrees. (Aaron Bentley)
- Help topics can now be loaded from files. (Ian Clatworthy, Alexander Belchenko)
- get_parent_map now always provides tuples as its output. (Aaron Bentley)
- Parent Providers should now implement get_parent_map returning a dictionary instead of get_parents returning a list. Graph.get_parents is now deprecated. (John Arbash Meinel, Robert Collins)
- Patience Diff now supports arbitrary python objects, as long as they support hash(). (John Arbash Meinel)
- Reduce selftest overhead to establish test names by memoization. (Vincent Ladeuil)
API BREAKS:
TESTING:
- Modules can now customise their tests by defining a load_tests attribute. pydoc bzrlib.tests.TestUtil.TestLoader.loadTestsFromModule for the documentation on this attribute. (Robert Collins)
- New helper function bzrlib.tests.condition_id_re which helps filter tests based on a regular expression search on the tests id. (Robert Collins)
- New helper function bzrlib.tests.condition_isinstance which helps filter tests based on class. (Robert Collins)
- New helper function bzrlib.tests.exclude_suite_by_condition which generalises the exclude_suite_by_re function. (Robert Collins)
- New helper function bzrlib.tests.filter_suite_by_condition which generalises the filter_suite_by_re function. (Robert Collins)
- New helper method bzrlib.tests.exclude_tests_by_re which gives a new TestSuite that does not contain tests from the input that matched a regular expression. (Robert Collins)
- New helper method bzrlib.tests.randomize_suite which returns a randomized copy of the input suite. (Robert Collins)
- New helper method bzrlib.tests.split_suite_by_re which splits a test suite into two according to a regular expression. (Robert Collins)
- Parametrize all http tests for the transport implementations, the http protocol versions (1.0 and 1.1) and the authentication schemes. (Vincent Ladeuil)
- The exclude_pattern and random_order parameters to the function bzrlib.tests.filter_suite_by_re have been deprecated. (Robert Collins)
- The method bzrlib.tests.sort_suite_by_re has been deprecated. It is replaced by the new helper methods added in this release. (Robert Collins)
DOCUMENTATION:
- More improvements and fixes to the User Guide. (Ian Clatworthy)
- Add information on cherrypicking/rebasing to the User Guide. (Ian Clatworthy)
- Improve bug tracker integration documentation. (Ian Clatworthy)
- Minor edits to Bazaar in five minutes from David Roberts and to the rebasing section of the User Guide from Aaron Bentley. (Ian Clatworthy)
CHANGES:
- If a traceback occurs, users are now asked to report the bug through Launchpad (https://bugs.launchpad.net/bzr/), rather than by mail to the mailing list. (Martin Pool)
BUGFIXES:
- Fix Makefile rules for doc generation. (Ian Clatworthy, #175207)
- Give more feedback during long http downloads by making readv deliver data as it arrives for urllib, and issue more requests for pycurl. High latency networks are better handled by urllib, the pycurl implementation give more feedback but also incur more latency. (Vincent Ladeuil, #173010)
- Implement _make_parents_provider on RemoteRepository, allowing generating bundles against branches on a smart server. (Andrew Bennetts, #147836)
DOCUMENTATION:
- Improved user guide. (Ian Clatworthy)
- The single-page quick reference guide is now available as a PDF. (Ian Clatworthy)
INTERNALS:
- readv urllib http implementation is now a real iterator above the underlying socket and deliver data as soon as it arrives. 'get' still wraps its output in a StringIO. (Vincent Ladeuil)
IMPROVEMENTS:
- Added a --coverage option to selftest. (Andrew Bennetts)
- Annotate merge (merge-type=weave) now supports cherrypicking. (Aaron Bentley)
- bzr commit now doesn't print the revision number twice. (Matt Nordhoff, #172612)
- New configuration option bugtracker_<tracker_abbrevation>_url to define locations of bug trackers that are not directly supported by bzr or a plugin. The URL will be treated as a template and {id} placeholders will be replaced by specific bug IDs. (Lukáš Lalinský)
- Support logging single merge revisions with short and line log formatters. (Kent Gibson)
- User Guide enhanced with suggested readability improvements from Matt Revell and corrections from John Arbash Meinel. (Ian Clatworthy)
- Quick Start Guide renamed to Quick Start Card, moved down in the catalog, provided in pdf and png format and updated to refer to send instead of bundle. (Ian Clatworthy, #165080)
- switch can now be used on heavyweight checkouts as well as lightweight ones. After switching a heavyweight checkout, the local branch is a mirror/cache of the new bound branch and uncommitted changes in the working tree are merged. As a safety check, if there are local commits in a checkout which have not been committed to the previously bound branch, then switch fails unless the --force option is given. This option is now also required if the branch a lightweight checkout is pointing to has been moved. (Ian Clatworthy)
INTERNALS:
- New -Dhttp debug option reports http connections, requests and responses. (Vincent Ladeuil)
- New -Dmerge debug option, which emits merge plans for merge-type=weave.
BUGFIXES:
- Better error message when running bzr cat on a non-existant branch. (Lukáš Lalinský, #133782)
- Catch OSError 17 (file exists) in final phase of tree transform and show filename to user. (Alexander Belchenko, #111758)
- Catch ShortReadvErrors while using pycurl. Also make readv more robust by allowing multiple GET requests to be issued if too many ranges are required. (Vincent Ladeuil, #172701)
- Check for missing basis texts when fetching from packs to packs. (John Arbash Meinel, #165290)
- Fall back to showing e-mail in log --short/--line if the committer/author has only e-mail. (Lukáš Lalinský, #157026)
API BREAKS:
- Deprecate not passing a location argument to commit reporters' started methods. (Matt Nordhoff)
NOTES WHEN UPGRADING:
The default repository format is now pack-0.92. This default is used when creating new repositories with init and init-repo, and when branching over bzr+ssh or bzr+hpss. (See https://bugs.launchpad.net/bugs/164626)
This format can be read and written by Bazaar 0.92 and later, and data can be transferred to and from older formats.
To upgrade, please reconcile your repository (bzr reconcile), and then upgrade (bzr upgrade).
pack-0.92 offers substantially better scaling and performance than the previous knits format. Some operations are slower where the code already had bad scaling characteristics under knits, the pack format makes such operations more visible as part of being more scalable overall. We will correct such operations over the coming releases and encourage the filing of bugs on any operation which you observe to be slower in a packs repository. One particular case that we do not intend to fix is pulling data from a pack repository into a knit repository over a high latency link; downgrading such data requires reinsertion of the file texts, and this is a classic space/time tradeoff. The current implementation is conservative on memory usage because we need to support converting data from any tree without problems. (Robert Collins, Martin Pool, #164476)
CHANGES:
- Disable detection of plink.exe as possible ssh vendor. Plink vendor still available if user selects it explicitly with BZR_SSH environment variable. (Alexander Belchenko, workaround for bug #107593)
- The pack format is now accessible as "pack-0.92", or "pack-0.92-subtree" to enable the subtree functions (for example, for bzr-svn). See http://doc.bazaar-vcs.org/latest/developer/packrepo.html (Martin Pool)
FEATURES:
- New authentication.conf file holding the password or other credentials for remote servers. This can be used for ssh, sftp, smtp and other supported transports. (Vincent Ladeuil)
- New rich-root and rich-root-pack formats, recording the same data about tree roots that's recorded for all other directories. (Aaron Bentley, #164639)
- pack-0.92 repositories can now be reconciled. (Robert Collins, #154173)
- switch command added for changing the branch a lightweight checkout is associated with and updating the tree to reflect the latest content accordingly. This command was previously part of the BzrTools plug-in. (Ian Clatworthy, Aaron Bentley, David Allouche)
- reconfigure command can now convert branches, trees, or checkouts to lightweight checkouts. (Aaron Bentley)
PERFORMANCE:
- Commit updates the state of the working tree via a delta rather than supplying entirely new basis trees. For commit of a single specified file this reduces the wall clock time for commit by roughly a 30%. (Robert Collins, Martin Pool)
- Commit with many automatically found deleted paths no longer performs linear scanning for the children of those paths during inventory iteration. This should fix commit performance blowing out when many such paths occur during commit. (Robert Collins, #156491)
- Fetch with pack repositories will no longer read the entire history graph. (Robert Collins, #88319)
- Revert takes out an appropriate lock when reverting to a basis tree, and does not read the basis inventory twice. (Robert Collins)
- Diff does not require an inventory to be generated on dirstate trees. (Aaron Bentley, #149254)
- New annotate merge (--merge-type=weave) implementation is fast on versionedfiles withough cached annotations, e.g. pack-0.92. (Aaron Bentley)
IMPROVEMENTS:
- bzr merge now warns when it encounters a criss-cross merge. (Aaron Bentley)
- bzr send now doesn't require the target e-mail address to be specified on the command line if an interactive e-mail client is used. (Lukáš Lalinský)
- bzr tags now prints the revision number for each tag, instead of the revision id, unless --show-ids is passed. In addition, tags can be sorted chronologically instead of lexicographically with --sort=time. (Adeodato Simó, #120231)
- Windows standalone version of bzr is able to load system-wide plugins from "plugins" subdirectory in installation directory. In addition standalone installer write to the registry (HKLMSOFTWAREBazaar) useful info about paths and bzr version. (Alexander Belchenko, #129298)
DOCUMENTATION:
BUG FIXES:
- A progress bar has been added for knitpack -> knitpack fetching. (Robert Collins, #157789, #159147)
- Branching from a branch via smart server now preserves the repository format. (Andrew Bennetts, #164626)
- commit is now able to invoke an external editor in a non-ascii directory. (Daniel Watkins, #84043)
- Catch connection errors for ftp. (Vincent Ladeuil, #164567)
- check no longer reports spurious unreferenced text versions. (Robert Collins, John A Meinel, #162931, #165071)
- Conflicts are now resolved recursively by revert. (Aaron Bentley, #102739)
- Detect invalid transport reuse attempts by catching invalid URLs. (Vincent Ladeuil, #161819)
- Deleting a file without removing it shows a correct diff, not a traceback. (Aaron Bentley)
- Do no use timeout in HttpServer anymore. (Vincent Ladeuil, #158972).
- Don't catch the exceptions related to the http pipeline status before retrying an http request or some programming errors may be masked. (Vincent Ladeuil, #160012)
- Fix bzr rm to not delete modified and ignored files. (Lukáš Lalinský, #172598)
- Fix exception when revisionspec contains merge revisons but log formatter doesn't support merge revisions. (Kent Gibson, #148908)
- Fix exception when ScopeReplacer is assigned to before any members have been retrieved. (Aaron Bentley)
- Fix multiple connections during checkout --lightweight. (Vincent Ladeuil, #159150)
- Fix possible error in insert_data_stream when copying between pack repositories over bzr+ssh or bzr+http. KnitVersionedFile.get_data_stream now makes sure that requested compression parents are sent before any delta hunks that depend on them. (Martin Pool, #164637)
- Fix typo in limiting offsets coalescing for http, leading to whole files being downloaded instead of parts. (Vincent Ladeuil, #165061)
- FTP server errors don't error in the error handling code. (Robert Collins, #161240)
- Give a clearer message when a pull fails because the source needs to be reconciled. (Martin Pool, #164443)
- It is clearer when a plugin cannot be loaded because of its name, and a suggestion for an acceptable name is given. (Daniel Watkins, #103023)
- Leave port as None in transport objects if user doesn't specify a port in urls. (vincent Ladeuil, #150860)
- Make sure Repository.fetch(self) is properly a no-op for all Repository implementations. (John Arbash Meinel, #158333)
- Mark .bzr directories as "hidden" on Windows. (Alexander Belchenko, #71147)
- merge --uncommitted can now operate on a single file. (Aaron Bentley, Lukáš Lalinský, #136890)
- Obsolete packs are now cleaned up by pack and autopack operations. (Robert Collins, #153789)
- Operations pulling data from a smart server where the underlying repositories are not both annotated/both unannotated will now work. (Robert Collins, #165304).
- Reconcile now shows progress bars. (Robert Collins, #159351)
- RemoteBranch was not initializing self._revision_id_to_revno_map properly. (John Arbash Meinel, #162486)
- Removing an already-removed file reports the file does not exist. (Daniel Watkins, #152811)
- Rename on Windows is able to change filename case. (Alexander Belchenko, #77740)
- Return error instead of a traceback for bzr log -r0. (Kent Gibson, #133751)
- Return error instead of a traceback when bzr is unable to create symlink on some platforms (e.g. on Windows). (Alexander Belchenko, workaround for #81689)
- Revert doesn't crash when restoring a single file from a deleted directory. (Aaron Bentley)
- Stderr output via logging mechanism now goes through encoded wrapper and no more uses utf-8, but terminal encoding instead. So all unicode strings now should be readable in non-utf-8 terminal. (Alexander Belchenko, #54173)
- The error message when move --after should be used makes how to do so clearer. (Daniel Watkins, #85237)
- Unicode-safe output from bzr info. The output will be encoded using the terminal encoding and unrepresentable characters will be replaced by '?'. (Lukáš Lalinský, #151844)
- Working trees are no longer created when pushing into a local no-trees repo. (Daniel Watkins, #50582)
- Upgrade util/configobj to version 4.4.0. (Vincent Ladeuil, #151208).
- Wrap medusa ftp test server as an FTPServer feature. (Vincent Ladeuil, #157752)
API BREAKS:
- osutils.backup_file is deprecated. Actually it's not used in bzrlib during very long time. (Alexander Belchenko)
- The return value of VersionedFile.iter_lines_added_or_present_in_versions has been changed. Previously it was an iterator of lines, now it is an iterator of (line, version_id) tuples. This change has been made to aid reconcile and fetch operations. (Robert Collins)
- bzrlib.repository.get_versioned_file_checker is now private. (Robert Collins)
- The Repository format registry default has been removed; it was previously obsoleted by the bzrdir format default, which implies a default repository format. (Martin Pool)
INTERNALS:
- Added ContainerSerialiser and ContainerPushParser to bzrlib.pack. These classes provide more convenient APIs for generating and parsing containers from streams rather than from files. (Andrew Bennetts)
- New module lru_cache providing a cache for use by tasks that need semi-random access to large amounts of data. (John A Meinel)
- InventoryEntry.diff is now deprecated. Please use diff.DiffTree instead.
TESTING:
CHANGES:
- New uninstaller on Win32. (Alexander Belchenko)
NOTES WHEN UPGRADING:
CHANGES:
- bzr now returns exit code 4 if an internal error occurred, and 3 if a normal error occurred. (Martin Pool)
- pull, merge and push will no longer silently correct some repository index errors that occured as a result of the Weave disk format. Instead the reconcile command needs to be run to correct those problems if they exist (and it has been able to fix most such problems since bzr 0.8). Some new problems have been identified during this release and you should run bzr check once on every repository to see if you need to reconcile. If you cannot pull or merge from a remote repository due to mismatched parent errors - a symptom of index errors - you should simply take a full copy of that remote repository to a clean directory outside any local repositories, then run reconcile on it, and finally pull from it locally. (And naturally email the repositories owner to ask them to upgrade and run reconcile). (Robert Collins)
FEATURES:
- New knitpack-experimental repository format. This is interoperable with the dirstate-tags format but uses a smarter storage design that greatly speeds up many operations, both local and remote. This new format can be used as an option to the init, init-repository and upgrade commands. See http://doc.bazaar-vcs.org/0.92/developers/knitpack.html for further details. (Robert Collins)
- For users of bzr-svn (and those testing the prototype subtree support) that wish to try packs, a new knitpack-subtree-experimental format has also been added. This is interoperable with the dirstate-subtrees format. (Robert Collins)
- New reconfigure command. (Aaron Bentley)
- New revert --forget-merges command, which removes the record of a pending merge without affecting the working tree contents. (Martin Pool)
- New bzr_remote_path configuration variable allows finer control of remote bzr locations than BZR_REMOTE_PATH environment variable. (Aaron Bentley)
- New launchpad-login command to tell Bazaar your Launchpad user ID. This can then be used by other functions of the Launchpad plugin. (James Henstridge)
PERFORMANCE:
- Commit in quiet mode is now slightly faster as the information to output is no longer calculated. (Ian Clatworthy)
- Commit no longer checks for new text keys during insertion when the revision id was deterministically unique. (Robert Collins)
- Committing a change which is not a merge and does not change the number of files in the tree is faster by utilising the data about whether files are changed to determine if the tree is unchanged rather than recalculating it at the end of the commit process. (Robert Collins)
- Inventory serialisation no longer double-sha's the content. (Robert Collins)
- Knit text reconstruction now avoids making copies of the lines list for interim texts when building a single text. The new apply_delta method on KnitContent aids this by allowing modification of the revision id such objects represent. (Robert Collins)
- Pack indices are now partially parsed for specific key lookup using a bisection approach. (Robert Collins)
- Partial commits are now approximately 40% faster by walking over the unselected current tree more efficiently. (Robert Collins)
- XML inventory serialisation takes 20% less time while being stricter about the contents. (Robert Collins)
- Graph heads() queries have been fixed to no longer access all history unnecessarily. (Robert Collins)
IMPROVEMENTS:
- bzr+https:// smart server across https now supported. (John Ferlito, Martin Pool, #128456)
- Mutt is now a supported mail client; set mail_client=mutt in your bazaar.conf and send will use mutt. (Keir Mierle)
- New option -c/--change for merge command for cherrypicking changes from one revision. (Alexander Belchenko, #141368)
- Show encodings, locale and list of plugins in the traceback message. (Martin Pool, #63894)
- Experimental directory formats can now be marked with experimental = True during registration. (Ian Clatworthy)
DOCUMENTATION:
- New Bazaar in Five Minutes guide. (Matthew Revell)
- The hooks reference documentation is now converted to html as expected. (Ian Clatworthy)
BUG FIXES:
- Connection error reporting for the smart server has been fixed to display a user friendly message instead of a traceback. (Ian Clatworthy, #115601)
- Make sure to use O_BINARY when opening files to check their sha1sum. (Alexander Belchenko, John Arbash Meinel, #153493)
- Fix a problem with Win32 handling of the executable bit. (John Arbash Meinel, #149113)
- bzr+ssh:// and sftp:// URLs that do not specify ports explicitly no longer assume that means port 22. This allows people using OpenSSH to override the default port in their ~/.ssh/config if they wish. This fixes a bug introduced in bzr 0.91. (Andrew Bennetts, #146715)
- Commands reporting exceptions can now be profiled and still have their data correctly dumped to a file. For example, a bzr commit with no changes still reports the operation as pointless but doing so no longer throws away the profiling data if this command is run with --lsprof-file callgrind.out.ci say. (Ian Clatworthy)
- Fallback to ftp when paramiko is not installed and sftp can't be used for tests/commands so that the test suite is still usable without paramiko. (Vincent Ladeuil, #59150)
- Fix commit ordering in corner case. (Aaron Bentley, #94975)
- Fix long standing bug in partial commit when there are renames left in tree. (Robert Collins, #140419)
- Fix selftest semi-random noise during http related tests. (Vincent Ladeuil, #140614)
- Fix typo in ftp.py making the reconnection fail on temporary errors. (Vincent Ladeuil, #154259)
- Fix failing test by comparing real paths to cover the case where the TMPDIR contains a symbolic link. (Vincent Ladeuil, #141382).
- Fix log against smart server branches that don't support tags. (James Westby, #140615)
- Fix pycurl http implementation by defining error codes from pycurl instead of relying on an old curl definition. (Vincent Ladeuil, #147530)
- Fix 'unprintable error' message when displaying BzrCheckError and some other exceptions on Python 2.5. (Martin Pool, #144633)
- Fix Inventory.copy() and add test for it. (Jelmer Vernooij)
- Handles default value for ListOption in cmd_commit. (Vincent Ladeuil, #140432)
- HttpServer and FtpServer need to be closed properly or a listening socket will remain opened. (Vincent Ladeuil, #140055)
- Monitor the .bzr directory created in the top level test directory to detect leaking tests. (Vincent Ladeuil, #147986)
- The basename, not the full path, is now used when checking whether the profiling dump file begins with callgrind.out or not. This fixes a bug reported by Aaron Bentley on IRC. (Ian Clatworthy)
- Trivial fix for invoking command reconfigure without arguments. (Rob Weir, #141629)
- WorkingTree.rename_one will now raise an error if normalisation of the new path causes bzr to be unable to access the file. (Robert Collins)
- Correctly detect a NoSuchFile when using a filezilla server. (Gary van der Merwe)
API BREAKS:
- bzrlib.index.GraphIndex now requires a size parameter to the constructor, for enabling bisection searches. (Robert Collins)
- CommitBuilder.record_entry_contents now requires the root entry of a tree be supplied to it, previously failing to do so would trigger a deprecation warning. (Robert Collins)
- KnitVersionedFile.add* will no longer cache added records even when enable_cache() has been called - the caching feature is now exclusively for reading existing data. (Robert Collins)
- ReadOnlyLockError is deprecated; LockFailed is usually more appropriate. (Martin Pool)
- Removed bzrlib.transport.TransportLogger - please see the new trace+ transport instead. (Robert Collins)
- Removed previously deprecated varargs interface to TestCase.run_bzr and deprecated methods TestCase.capture and TestCase.run_bzr_captured. (Martin Pool)
- Removed previous deprecated basis_knit parameter to the KnitVersionedFile constructor. (Robert Collins)
- Special purpose method TestCase.run_bzr_decode is moved to the test_non_ascii class that needs it. (Martin Pool)
- The class bzrlib.repofmt.knitrepo.KnitRepository3 has been folded into KnitRepository by parameters to the constructor. (Robert Collins)
- The VersionedFile interface now allows content checks to be bypassed by supplying check_content=False. This saves nearly 30% of the minimum cost to store a version of a file. (Robert Collins)
- Tree's with bad state such as files with no length or sha will no longer be silently accepted by the repository XML serialiser. To serialise inventories without such data, pass working=True to write_inventory. (Robert Collins)
- VersionedFile.fix_parents has been removed as a harmful API. VersionedFile.join will no longer accept different parents on either side of a join - it will either ignore them, or error, depending on the implementation. See notes when upgrading for more information. (Robert Collins)
INTERNALS:
- bzrlib.transport.Transport.put_file now returns the number of bytes put by the method call, to allow avoiding stat-after-write or housekeeping in callers. (Robert Collins)
- bzrlib.xml_serializer.Serializer is now responsible for checking that mandatory attributes are present on serialisation and deserialisation. This fixes some holes in API usage and allows better separation between physical storage and object serialisation. (Robert Collins)
- New class bzrlib.errors.InternalBzrError which is just a convenient shorthand for deriving from BzrError and setting internal_error = True. (Robert Collins)
- New method bzrlib.mutabletree.update_to_one_parent_via_delta for moving the state of a parent tree to a new version via a delta rather than a complete replacement tree. (Robert Collins)
- New method bzrlib.osutils.minimum_path_selection useful for removing duplication from user input, when a user mentions both a path and an item contained within that path. (Robert Collins)
- New method bzrlib.repository.Repository.is_write_locked useful for determining if a repository is write locked. (Robert Collins)
- New method on bzrlib.tree.Tree path_content_summary provides a tuple containing the key information about a path for commit processing to complete. (Robert Collins)
- New method on xml serialisers, write_inventory_to_lines, which matches the API used by knits for adding content. (Robert Collins)
- New module bzrlib.bisect_multi with generic multiple-bisection-at-once logic, currently only available for byte-based lookup (bisect_multi_bytes). (Robert Collins)
- New helper bzrlib.tuned_gzip.bytes_to_gzip which takes a byte string and returns a gzipped version of the same. This is used to avoid a bunch of api friction during adding of knit hunks. (Robert Collins)
- New parameter on bzrlib.transport.Transport.readv adjust_for_latency which changes readv from returning strictly the requested data to inserted return larger ranges and in forward read order to reduce the effect of network latency. (Robert Collins)
- New parameter yield_parents on Inventory.iter_entries_by_dir which causes the parents of a selected id to be returned recursively, so all the paths from the root down to each element of selected_file_ids are returned. (Robert Collins)
- Knit joining has been enhanced to support plain to annotated conversion and annotated to plain conversion. (Ian Clatworthy)
- The CommitBuilder method record_entry_contents now returns summary information about the effect of the commit on the repository. This tuple contains an inventory delta item if the entry changed from the basis, and a boolean indicating whether a new file graph node was recorded. (Robert Collins)
- The python path used in the Makefile can now be overridden. (Andrew Bennetts, Ian Clatworthy)
TESTING:
- New transport implementation trace+ which is useful for testing, logging activity taken to its _activity attribute. (Robert Collins)
- When running bzr commands within the test suite, internal exceptions are not caught and reported in the usual way, but rather allowed to propagate up and be visible to the test suite. A new API run_bzr_catch_user_errors makes this behavior available to other users. (Martin Pool)
- New method TestCase.call_catch_warnings for testing methods that raises a Python warning. (Martin Pool)
BUG FIXES:
- Print a warning instead of aborting the python setup.py install process if building of a C extension is not possible. (Lukáš Lalinský, Alexander Belchenko)
- Fix commit ordering in corner case (Aaron Bentley, #94975)
- Fix ''bzr info bzr://host/'' and other operations on ''bzr://' URLs with an implicit port. We were incorrectly raising PathNotChild due to inconsistent treatment of the ''_port'' attribute on the Transport object. (Andrew Bennetts, #133965)
- Make RemoteRepository.sprout cope gracefully with servers that don't support the Repository.tarball request. (Andrew Bennetts)
- Replaced incorrect tarball for previous release; a debug statement was left in bzrlib/remote.py.
CHANGES:
- The default branch and repository format has changed to dirstate-tags, so tag commands are active by default. This format is compatible with Bazaar 0.15 and later. This incidentally fixes bug #126141. (Martin Pool)
- --quiet or -q is no longer a global option. If present, it must now appear after the command name. Scripts doing things like bzr -q missing need to be rewritten as bzr missing -q. (Ian Clatworthy)
FEATURES:
New option --author in bzr commit to specify the author of the change, if it's different from the committer. bzr log and bzr annotate display the author instead of the committer. (Lukáš Lalinský)
In addition to global options and command specific options, a set of standard options are now supported. Standard options are legal for all commands. The initial set of standard options are:
- --help or -h - display help message
- --verbose or -v - display additional information
- --quiet or -q - only output warnings and errors.
Unlike global options, standard options can be used in aliases and may have command-specific help. (Ian Clatworthy)
Verbosity level processing has now been unified. If --verbose or -v is specified on the command line multiple times, the verbosity level is made positive the first time then increased. If --quiet or -q is specified on the command line multiple times, the verbosity level is made negative the first time then decreased. To get the default verbosity level of zero, either specify none of the above , --no-verbose or --no-quiet. Note that most commands currently ignore the magnitude of the verbosity level but do respect quiet vs normal vs verbose when generating output. (Ian Clatworthy)
Branch.hooks now supports pre_commit hook. The hook's signature is documented in BranchHooks constructor. (Nam T. Nguyen, #102747)
New Repository.stream_knit_data_for_revisions request added to the network protocol for greatly reduced roundtrips when retrieving a set of revisions. (Andrew Bennetts)
BUG FIXES:
- bzr plugins now lists the version number for each plugin in square brackets after the path. (Robert Collins, #125421)
- Pushing, pulling and branching branches with subtree references was not copying the subtree weave, preventing the file graph from being accessed and causing errors in commits in clones. (Robert Collins)
- Suppress warning "integer argument expected, got float" from Paramiko, which sometimes caused false test failures. (Martin Pool)
- Fix bug in bundle 4 that could cause attempts to write data to wrong versionedfile. (Aaron Bentley)
- Diffs generated using "diff -p" no longer break the patch parser. (Aaron Bentley)
- get_transport treats an empty possible_transports list the same as a non- empty one. (Aaron Bentley)
- patch verification for merge directives is reactivated, and works with CRLF and CR files. (Aaron Bentley)
- Accept ..as a path in revision specifiers. This fixes for example "-r branch:..other-branch" on Windows. (Lukáš Lalinský)
- BZR_PLUGIN_PATH may now contain trailing slashes. (Blake Winton, #129299)
- man page no longer lists hidden options (#131667, Aaron Bentley)
- uncommit --help now explains the -r option adequately. (Daniel Watkins, #106726)
- Error messages are now better formatted with parameters (such as filenames) quoted when necessary. This avoids confusion when directory names ending in a '.' at the end of messages were confused with a full stop that may or not have been there. (Daniel Watkins, #129791)
- Fix status FILE -r X..Y. (Lukáš Lalinský)
- If a particular command is an alias, help will show the alias instead of claiming there is no help for said alias. (Daniel Watkins, #133548)
- TreeTransform-based operations, like pull, merge, revert, and branch, now roll back if they encounter an error. (Aaron Bentley, #67699)
- bzr commit now exits cleanly if a character unsupported by the current encoding is used in the commit message. (Daniel Watkins, #116143)
- bzr send uses default values for ranges when only half of an elipsis is specified ("-r..5" or "-r5.."). (#61685, Aaron Bentley)
- Avoid trouble when Windows ssh calls itself 'plink' but no plink binary is present. (Martin Albisetti, #107155)
- bzr remove should remove clean subtrees. Now it will remove (without needing --force) subtrees that contain no files with text changes or modified files. With --force it removes the subtree regardless of text changes or unknown files. Directories with renames in or out (but not changed otherwise) will now be removed without needing --force. Unknown ignored files will be deleted without needing --force. (Marius Kruger, #111665)
- When two plugins conflict, the source of both the losing and now the winning definition is shown. (Konstantin Mikhaylov, #5454)
- When committing to a branch, the location being committed to is displayed. (Daniel Watkins, #52479)
- bzr --version takes care about encoding of stdout, especially when output is redirected. (Alexander Belchenko, #131100)
- Prompt for an ftp password if none is provided. (Vincent Ladeuil, #137044)
- Reuse bound branch associated transport to avoid multiple connections. (Vincent Ladeuil, #128076, #131396)
- Overwrite conflicting tags by push and pull if the --overwrite option is specified. (Lukáš Lalinský, #93947)
- In checkouts, tags are copied into the master branch when created, changed or deleted, and are copied into the checkout when it is updated. (Martin Pool, #93856, #93860)
- Print a warning instead of aborting the python setup.py install process if building of a C extension is not possible. (Lukáš Lalinský, Alexander Belchenko)
IMPROVEMENTS:
- Add the option "--show-diff" to the commit command in order to display the diff during the commit log creation. (Goffredo Baroncelli)
- pull and merge are much faster at installing bundle format 4. (Aaron Bentley)
- pull -v no longer includes deltas, making it much faster. (Aaron Bentley)
- send now sends the directive as an attachment by default. (Aaron Bentley, Lukáš Lalinský, Alexander Belchenko)
- Documentation updates (Martin Albisetti)
- Help on debug flags is now included in help global-options. (Daniel Watkins, #124853)
- Parameters passed on the command line are checked to ensure they are supported by the encoding in use. (Daniel Watkins)
- The compression used within the bzr repository has changed from zlib level 9 to the zlib default level. This improves commit performance with only a small increase in space used (and in some cases a reduction in space). (Robert Collins)
- Initial commit no longer SHAs files twice and now reuses the path rather than looking it up again, making it faster. (Ian Clatworthy)
- New option -c/--change for diff and status to show changes in one revision. (Lukáš Lalinský)
- If versioned files match a given ignore pattern, a warning is now given. (Daniel Watkins, #48623)
- bzr status now has -S as a short name for --short and -V as a short name for --versioned. These have been added to assist users migrating from Subversion: bzr status -SV is now like svn status -q. (Daniel Watkins, #115990)
- Added C implementation of PatienceSequenceMatcher, which is about 10x faster than the Python version. This speeds up commands that need file diffing, such as bzr commit or bzr diff. (Lukáš Lalinský)
- HACKING has been extended with a large section on core developer tasks. (Ian Clatworthy)
- Add branches and standalone-trees as online help topics and include them as Concepts within the User Reference. (Paul Moore, Ian Clatworthy)
- check can detect versionedfile parent references that are inconsistent with revision and inventory info, and reconcile can fix them. These faulty references were generated by 0.8-era releases, so repositories which were manipulated by old bzrs should be checked, and possibly reconciled ASAP. (Aaron Bentley, Andrew Bennetts)
API BREAKS:
- Branch.append_revision is removed altogether; please use Branch.set_last_revision_info instead. (Martin Pool)
- CommitBuilder now advertises itself as requiring the root entry to be supplied. This only affects foreign repository implementations which reuse CommitBuilder directly and have changed record_entry_contents to require that the root not be supplied. This should be precisely zero plugins affected. (Robert Collins)
- The add_lines methods on VersionedFile implementations has changed its return value to include the sha1 and length of the inserted text. This allows the avoidance of double-sha1 calculations during commit. (Robert Collins)
- Transport.should_cache has been removed. It was not called in the previous release. (Martin Pool)
TESTING:
- Tests may now raise TestNotApplicable to indicate they shouldn't be run in a particular scenario. (Martin Pool)
- New function multiply_tests_from_modules to give a simpler interface to test parameterization. (Martin Pool, Robert Collins)
- Transport.should_cache has been removed. It was not called in the previous release. (Martin Pool)
- NULL_REVISION is returned to indicate the null revision, not None. (Aaron Bentley)
- Use UTF-8 encoded StringIO for log tests to avoid failures on non-ASCII committer names. (Lukáš Lalinský)
INTERNALS:
- bzrlib.plugin.all_plugins has been deprecated in favour of bzrlib.plugin.plugins() which returns PlugIn objects that provide useful functionality for determining the path of a plugin, its tests, and its version information. (Robert Collins)
- Add the option user_encoding to the function 'show_diff_trees()' in order to move the user encoding at the UI level. (Goffredo Baroncelli)
- Add the function make_commit_message_template_encoded() and the function edit_commit_message_encoded() which handle encoded strings. This is done in order to mix the commit messages (which is a unicode string), and the diff which is a raw string. (Goffredo Baroncelli)
- CommitBuilder now defaults to using add_lines_with_ghosts, reducing overhead on non-weave repositories which don't require all parents to be present. (Robert Collins)
- Deprecated method find_previous_heads on bzrlib.inventory.InventoryEntry. This has been superseded by the use of parent_candidates and a separate heads check via the repository API. (Robert Collins)
- New trace function mutter_callsite will print out a subset of the stack to the log, which can be useful for gathering debug details. (Robert Collins)
- bzrlib.pack.ContainerWriter now tracks how many records have been added via a public attribute records_written. (Robert Collins)
- New method bzrlib.transport.Transport.get_recommended_page_size. This provides a hint to users of transports as to the reasonable minimum data to read. In principle this can take latency and bandwidth into account on a per-connection basis, but for now it just has hard coded values based on the url. (e.g. http:// has a large page size, file:// has a small one.) (Robert Collins)
- New method on bzrlib.transport.Transport open_write_stream allows incremental addition of data to a file without requiring that all the data be buffered in memory. (Robert Collins)
- New methods on bzrlib.knit.KnitVersionedFile: get_data_stream(versions), insert_data_stream(stream) and get_format_signature(). These provide some infrastructure for efficiently streaming the knit data for a set of versions over the smart protocol.
- Knits with no annotation cache still produce correct annotations. (Aaron Bentley)
- Three new methods have been added to bzrlib.trace: set_verbosity_level, get_verbosity_level and is_verbose. set_verbosity_level expects a numeric value: negative for quiet, zero for normal, positive for verbose. The size of the number can be used to determine just how quiet or verbose the application should be. The existing be_quiet and is_quiet routines have been integrated into this new scheme. (Ian Clatworthy)
- Options can now be delcared with a custom_callback parameter. If set, this routine is called after the option is processed. This feature is now used by the standard options verbose and quiet so that setting one implicitly resets the other. (Ian Clatworthy)
- Rather than declaring a new option from scratch in order to provide custom help, a centrally registered option can be decorated using the new bzrlib.Option.custom_help routine. In particular, this routine is useful when declaring better help for the verbose and quiet standard options as the base definition of these is now more complex than before thanks to their use of a custom callback. (Ian Clatworthy)
- Tree._iter_changes(specific_file=[]) now iterates through no files, instead of iterating through all files. None is used to iterate through all files. (Aaron Bentley)
- WorkingTree.revert() now accepts None to revert all files. The use of [] to revert all files is deprecated. (Aaron Bentley)
IMPROVEMENTS:
- Documentation is now organized into multiple directories with a level added for different languages or locales. Added the Mini Tutorial and Quick Start Summary (en) documents from the Wiki, improving the content and readability of the former. Formatted NEWS as Release Notes complete with a Table of Conents, one heading per release. Moved the Developer Guide into the main document catalog and provided a link from the developer document catalog back to the main one. (Ian Clatworthy, Sabin Iacob, Alexander Belchenko)
API CHANGES:
- The static convenience method BzrDir.create_repository is deprecated. Callers should instead create a BzrDir instance and call create_repository on that. (Martin Pool)
BUGFIXES:
- bzr init should connect to the remote location one time only. We have been connecting several times because we forget to pass around the Transport object. This modifies BzrDir.create_branch_convenience, so that we can give it the Transport we already have. (John Arbash Meinel, Vincent Ladeuil, #111702)
- Get rid of sftp connection cache (get rid of the FTP one too). (Vincent Ladeuil, #43731)
- bzr branch {local|remote} remote don't try to create a working tree anymore. (Vincent Ladeuil, #112173)
- All identified multiple connections for a single bzr command have been fixed. See bzrlib/tests/commands directory. (Vincent Ladeuil)
- bzr rm now does not insist on --force to delete files that have been renamed but not otherwise modified. (Marius Kruger, #111664)
- bzr selftest --bench no longer emits deprecation warnings (Lukáš Lalinský)
- bzr status now honours FILE parameters for conflict lists (Aaron Bentley, #127606)
- bzr checkout now honours -r when reconstituting a working tree. It also honours -r 0. (Aaron Bentley, #127708)
- bzr add * no more fails on Windows if working tree contains non-ascii file names. (Kuno Meyer, #127361)
- allow easy_install bzr runs without fatal errors. (Alexander Belchenko, #125521)
- Graph._filter_candidate_lca does not raise KeyError if a candidate is eliminated just before it would normally be examined. (Aaron Bentley)
- SMTP connection failures produce a nice message, not a traceback. (Aaron Bentley)
IMPROVEMENTS:
- Don't show "dots" progress indicators when run non-interactively, such as from cron. (Martin Pool)
- info now formats locations more nicely and lists "submit" and "public" branches (Aaron Bentley)
- New pack command that will trigger database compression within the repository (Robert Collins)
- Implement _KnitIndex._load_data in a pyrex extension. The pyrex version is approximately 2-3x faster at parsing a .kndx file. Which yields a measurable improvement for commands which have to read from the repository, such as a 1s => 0.75s improvement in bzr diff when there are changes to be shown. (John Arbash Meinel)
- Merge is now faster. Depending on the scenario, it can be more than 2x faster. (Aaron Bentley)
- Give a clearer warning, and allow python setup.py install to succeed even if pyrex is not available. (John Arbash Meinel)
- DirState._read_dirblocks now has an optional Pyrex implementation. This improves the speed of any command that has to read the entire DirState. (diff, status, etc, improve by about 10%). bisect_dirblocks has also been improved, which helps all _get_entry type calls (whenever we are searching for a particular entry in the in-memory DirState). (John Arbash Meinel)
- bzr pull and bzr push no longer do a complete walk of the branch revision history for ui display unless -v is supplied. (Robert Collins)
- bzr log -rA..B output shifted to the left margin if the log only contains merge revisions. (Kent Gibson)
- The plugins command is now public with improved help. (Ian Clatworthy)
- New bundle and merge directive formats are faster to generate, and
- Annotate merge now works when there are local changes. (Aaron Bentley)
- Commit now only shows the progress in terms of directories instead of entries. (Ian Clatworthy)
- Fix KnitRepository.get_revision_graph to not request the graph 2 times. This makes get_revision_graph 2x faster. (John Arbash Meinel)
- Fix VersionedFile.get_graph() to avoid using set.difference_update(other), which has bad scaling when other is large. This improves VF.get_graph([version_id]) for a 12.5k graph from 2.9s down to 200ms. (John Arbash Meinel)
- The --lsprof-file option now generates output for KCacheGrind if the file starts with callgrind.out. This matches the default file filtering done by KCacheGrind's Open Dialog. (Ian Clatworthy)
- Fix bzr update to avoid an unnecessary branch.get_master_branch call, which avoids 1 extra connection to the remote server. (Partial fix for #128076, John Arbash Meinel)
- Log errors from the smart server in the trace file, to make debugging test failures (and live failures!) easier. (Andrew Bennetts)
- The HTML version of the man page has been superceded by a more comprehensive manual called the Bazaar User Reference. This manual is completed generated from the online help topics. As part of this change, limited reStructuredText is now explicitly supported in help topics and command help with 'unnatural' markup being removed prior to display by the online help or inclusion in the man page. (Ian Clatworthy)
- HTML documentation now use files extension *.html (Alexander Belchenko)
- The cache of ignore definitions is now cleared in WorkingTree.unlock() so that changes to .bzrignore aren't missed. (#129694, Daniel Watkins)
- bzr selftest --strict fails if there are any missing features or expected test failures. (Daniel Watkins, #111914)
- Link to registration survey added to README. (Ian Clatworthy)
- Windows standalone installer show link to registration survey when installation finished. (Alexander Belchenko)
LIBRARY API BREAKS:
- Deprecated dictionary bzrlib.option.SHORT_OPTIONS removed. Options are now required to provide a help string and it must comply with the style guide by being one or more sentences with an initial capital and final period. (Martin Pool)
- KnitIndex.get_parents now returns tuples. (Robert Collins)
- Ancient unused Repository.text_store attribute has been removed. (Robert Collins)
- The bzrlib.pack interface has changed to use tuples of bytestrings rather than just bytestrings, making it easier to represent multiple element names. As this interface was not used by any internal facilities since it was introduced in 0.18 no API compatibility is being preserved. The serialised form of these packs is identical with 0.18 when a single element tuple is in use. (Robert Collins)
INTERNALS:
- merge now uses iter_changes to calculate changes, which makes room for future performance increases. It is also more consistent with other operations that perform comparisons, and reduces reliance on Tree.inventory. (Aaron Bentley)
- Refactoring of transport classes connected to a remote server. ConnectedTransport is a new class that serves as a basis for all transports needing to connect to a remote server. transport.split_url have been deprecated, use the static method on the object instead. URL tests have been refactored too. (Vincent Ladeuil)
- Better connection sharing for ConnectedTransport objects. transport.get_transport() now accepts a 'possible_transports' parameter. If a newly requested transport can share a connection with one of the list, it will. (Vincent Ladeuil)
- Most functions now accept bzrlib.revision.NULL_REVISION to indicate the null revision, and consider using None for this purpose deprecated. (Aaron Bentley)
- New index module with abstract index functionality. This will be used during the planned changes in the repository layer. Currently the index layer provides a graph aware immutable index, a builder for the same index type to allow creating them, and finally a composer for such indices to allow the use of many indices in a single query. The index performance is not optimised, however the API is stable to allow development on top of the index. (Robert Collins)
- bzrlib.dirstate.cmp_by_dirs can be used to compare two paths by their directory sections. This is equivalent to comparing path.split('/'), only without having to split the paths. This has a Pyrex implementation available. (John Arbash Meinel)
- New transport decorator 'unlistable+' which disables the list_dir functionality for testing.
- Deprecated change_entry in transform.py. (Ian Clatworthy)
- RevisionTree.get_weave is now deprecated. Tree.plan_merge is now used for performing annotate-merge. (Aaron Bentley)
- New EmailMessage class to create email messages. (Adeodato Simó)
- Unused functions on the private interface KnitIndex have been removed. (Robert Collins)
- New knit.KnitGraphIndex which provides a KnitIndex layered on top of a index.GraphIndex. (Robert Collins)
- New knit.KnitVersionedFile.iter_parents method that allows querying the parents of many knit nodes at once, reducing round trips to the underlying index. (Robert Collins)
- Graph now has an is_ancestor method, various bits use it. (Aaron Bentley)
- The -Dhpss flag now includes timing information. As well as logging when a new connection is opened. (John Arbash Meinel)
- bzrlib.pack.ContainerWriter now returns an offset, length tuple to callers when inserting data, allowing generation of readv style access during pack creation, without needing a separate pass across the output pack to gather such details. (Robert Collins)
- bzrlib.pack.make_readv_reader allows readv based access to pack files that are stored on a transport. (Robert Collins)
- New Repository.has_same_location method that reports if two repository objects refer to the same repository (although with some risk of false negatives). (Andrew Bennetts)
- InterTree.compare now passes require_versioned on correctly. (Marius Kruger)
- New methods on Repository - start_write_group, commit_write_group, abort_write_group and is_in_write_group - which provide a clean hook point for transactional Repositories - ones where all the data for a fetch or commit needs to be made atomically available in one step. This allows the write lock to remain while making a series of data insertions. (e.g. data conversion). (Robert Collins)
- In bzrlib.knit the internal interface has been altered to use 3-tuples (index, pos, length) rather than two-tuples (pos, length) to describe where data in a knit is, allowing knits to be split into many files. (Robert Collins)
- bzrlib.knit._KnitData split into cache management and physical access with two access classes - _PackAccess and _KnitAccess defined. The former provides access into a .pack file, and the latter provides the current production repository form of .knit files. (Robert Collins)
TESTING:
- Remove selftest --clean-output, --numbered-dirs and --keep-output options, which are obsolete now that tests are done within directories in $TMPDIR. (Martin Pool)
- The SSH_AUTH_SOCK environment variable is now reset to avoid interaction with any running ssh agents. (Jelmer Vernooij, #125955)
- run_bzr_subprocess handles parameters the same way as run_bzr: either a string or a list of strings should be passed as the first parameter. Varargs-style parameters are deprecated. (Aaron Bentley)
BUGFIXES:
- Fix 'bzr add' crash under Win32 (Kuno Meyer)
BUGFIXES:
- Do not suppress pipe errors, etc. in non-display commands (Alexander Belchenko, #87178)
- Display a useful error message when the user requests to annotate a file that is not present in the specified revision. (James Westby, #122656)
- Commands that use status flags now have a reference to 'help status-flags'. (Daniel Watkins, #113436)
- Work around python-2.4.1 inhability to correctly parse the authentication header. (Vincent Ladeuil, #121889)
- Use exact encoding for merge directives. (Adeodato Simó, #120591)
- Fix tempfile permissions error in smart server tar bundling under Windows. (Martin _, #119330)
- Fix detection of directory entries in the inventory. (James Westby)
- Fix handling of http code 400: Bad Request When issuing too many ranges. (Vincent Ladeuil, #115209)
- Issue a CONNECT request when connecting to an https server via a proxy to enable SSL tunneling. (Vincent Ladeuil, #120678)
- Fix bzr log -r to support selecting merge revisions, both individually and as part of revision ranges. (Kent Gibson, #4663)
- Don't leave cruft behind when failing to acquire a lockdir. (Martin Pool, #109169)
- Don't use the '-f' strace option during tests. (Vincent Ladeuil, #102019).
- Warn when setting push_location to a value that will be masked by locations.conf. (Aaron Bentley, #122286)
- Fix commit ordering in corner case (Aaron Bentley, #94975)
- Make annotate behave in a non-ASCII world (Adeodato Simó).
IMPROVEMENTS:
The --lsprof-file option now dumps a text rendering of the profiling information if the filename ends in ".txt". It will also convert the profiling information to a format suitable for KCacheGrind if the output filename ends in ".callgrind". Fixes to the lsprofcalltree conversion process by Jean Paul Calderone and Itamar were also merged. See http://ddaa.net/blog/python/lsprof-calltree. (Ian Clatworthy)
info now defaults to non-verbose mode, displaying only paths and abbreviated format info. info -v displays all the information formerly displayed by info. (Aaron Bentley, Adeodato Simó)
bzr missing now has better option names --this and --other. (Elliot Murphy)
The internal weave-list command has become versionedfile-list, and now lists knits as well as weaves. (Aaron Bentley)
Automatic merge base selection uses a faster algorithm that chooses better bases in criss-cross merge situations (Aaron Bentley)
Progress reporting in commit has been improved. The various logical stages are now reported on as follows, namely:
- Collecting changes [Entry x/y] - Stage n/m
- Saving data locally - Stage n/m
- Uploading data to master branch - Stage n/m
- Updating the working tree - Stage n/m
- Running post commit hooks - Stage n/m
If there is no master branch, the 3rd stage is omitted and the total number of stages is adjusted accordingly.
Each hook that is run after commit is listed with a name (as hooks can be slow it is useful feedback). (Ian Clatworthy, Robert Collins)
Various operations that are now faster due to avoiding unnecessary topological sorts. (Aaron Bentley)
Make merge directives robust against broken bundles. (Aaron Bentley)
The lsprof filename note is emitted via trace.note(), not standard output. (Aaron Bentley)
bzrlib now exports explicit API compatibility information to assist library users and plugins. See the bzrlib.api module for details. (Robert Collins)
Remove unnecessary lock probes when acquiring a lockdir. (Martin Pool)
bzr --version now shows the location of the bzr log file, which is especially useful on Windows. (Martin Pool)
-D now supports hooks to get debug tracing of hooks (though its currently minimal in nature). (Robert Collins)
Long log format reports deltas on merge revisions. (John Arbash Meinel, Kent Gibson)
Make initial push over ftp more resilient. (John Arbash Meinel)
Print a summary of changes for update just like pull does. (Daniel Watkins, #113990)
Add a -Dhpss option to trace smart protocol requests and responses. (Andrew Bennetts)
LIBRARY API BREAKS:
- Testing cleanups - bzrlib.repository.RepositoryTestProviderAdapter has been moved to bzrlib.tests.repository_implementations; bzrlib.repository.InterRepositoryTestProviderAdapter has been moved to bzrlib.tests.interrepository_implementations; bzrlib.transport.TransportTestProviderAdapter has moved to bzrlib.tests.test_transport_implementations. bzrlib.branch.BranchTestProviderAdapter has moved to bzrlib.tests.branch_implementations. bzrlib.bzrdir.BzrDirTestProviderAdapter has moved to bzrlib.tests.bzrdir_implementations. bzrlib.versionedfile.InterVersionedFileTestProviderAdapter has moved to bzrlib.tests.interversionedfile_implementations. bzrlib.store.revision.RevisionStoreTestProviderAdapter has moved to bzrlib.tests.revisionstore_implementations. bzrlib.workingtree.WorkingTreeTestProviderAdapter has moved to bzrlib.tests.workingtree_implementations. These changes are an API break in the testing infrastructure only. (Robert Collins)
- Relocate TestCaseWithRepository to be more central. (Robert Collins)
- bzrlib.add.smart_add_tree will no longer perform glob expansion on win32. Callers of the function should do this and use the new MutableTree.smart_add method instead. (Robert Collins)
- bzrlib.add.glob_expand_for_win32 is now bzrlib.win32utils.glob_expand. (Robert Collins)
- bzrlib.add.FastPath is now private and moved to bzrlib.mutabletree._FastPath. (Robert Collins, Martin Pool)
- LockDir.wait removed. (Martin Pool)
- The SmartServer hooks API has changed for the server_started and server_stopped hooks. The first parameter is now an iterable of backing URLs rather than a single URL. This is to reflect that many URLs may map to the external URL of the server. E.g. the server interally may have a chrooted URL but also the local file:// URL will be at the same location. (Robert Collins)
INTERNALS:
- New SMTPConnection class to unify email handling. (Adeodato Simó)
- Fix documentation of BzrError. (Adeodato Simó)
- Make BzrBadParameter an internal error. (Adeodato Simó)
- Remove use of 'assert False' to raise an exception unconditionally. (Martin Pool)
- Give a cleaner error when failing to decode knit index entry. (Martin Pool)
- TreeConfig would mistakenly search the top level when asked for options from a section. It now respects the section argument and only searches the specified section. (James Westby)
- Improve make api-docs output. (John Arbash Meinel)
- Use os.lstat rather than os.stat for osutils.make_readonly and osutils.make_writeable. This makes the difftools plugin more robust when dangling symlinks are found. (Elliot Murphy)
- New -Dlock option to log (to ~/.bzr.log) information on when lockdirs are taken or released. (Martin Pool)
- bzrlib Hooks are now nameable using Hooks.name_hook. This allows a nicer UI when hooks are running as the current hook can be displayed. (Robert Collins)
- Transport.get has had its interface made more clear for ease of use. Retrieval of a directory must now fail with either 'PathError' at open time, or raise 'ReadError' on a read. (Robert Collins)
- New method _maybe_expand_globs on the Command class for dealing with unexpanded glob lists - e.g. on the win32 platform. This was moved from bzrlib.add._prepare_file_list. (Robert Collins)
- bzrlib.add.smart_add and bzrlib.add.smart_add_tree are now deprecated in favour of MutableTree.smart_add. (Robert Collins, Martin Pool)
- New method external_url on Transport for obtaining the url to hand to external processes. (Robert Collins)
- Teach windows installers to build pyrex/C extensions. (Alexander Belchenko)
TESTING:
- Removed the --keep-output option from selftest and clean up test directories as they're used. This reduces the IO load from running the test suite and cuts the time by about half. (Andrew Bennetts, Martin Pool)
- Add scenarios as a public attribute on the TestAdapter classes to allow modification of the generated scenarios before adaption and easier testing. (Robert Collins)
- New testing support class TestScenarioApplier which multiplies out a single teste by a list of supplied scenarios. (RobertCollins)
- Setting repository_to_test_repository on a repository_implementations test will cause it to be called during repository creation, allowing the testing of repository classes which are not based around the Format concept. For example a repository adapter can be tested in this manner, by altering the repository scenarios to include a scenario that sets this attribute during the test parameterisation in bzrlib.tests.repository.repository_implementations. (Robert Collins)
- Clean up many of the APIs for blackbox testing of Bazaar. The standard interface is now self.run_bzr. The command to run can be passed as either a list of parameters, a string containing the command line, or (deprecated) varargs parameters. (Martin Pool)
- The base TestCase now isolates tests from -D parameters by clearing debug.debug_flags and restores it afterwards. (Robert Collins)
- Add a relpath parameter to get_transport methods in test framework to avoid useless cloning. (Vincent Ladeuil, #110448)
BUGFIXES:
- Fix crash of commit due to wrong lookup of filesystem encoding. (Colin Watson, #120647)
- Revert logging just to stderr in commit as broke unicode filenames. (Aaron Bentley, Ian Clatworthy, #120930)
NOTES WHEN UPGRADING:
The kind() and is_executable() APIs on the WorkingTree interface no longer implicitly (read) locks and unlocks the tree. This might impact some plug-ins and tools using this part of the API. If you find an issue that may be caused by this change, please let us know, particularly the plug-in/tool maintainer. If encountered, the API fix is to surround kind() and is_executable() calls with lock_read() and unlock() like so:
work_tree.lock_read() try: kind = work_tree.kind(...) finally: work_tree.unlock()
- INTERNALS:
- Rework of LogFormatter API to provide beginning/end of log hooks and to encapsulate the details of the revision to be logged in a LogRevision object. In long log formats, merge revision ids are only shown when --show-ids is specified, and are labelled "revision-id:", as per mainline revisions, instead of "merged:". (Kent Gibson)
- New BranchBuilder API which allows the construction of particular histories quickly. Useful for testing and potentially other applications too. (Robert Collins)
IMPROVEMENTS:
- There are two new help topics, working-trees and repositories that attempt to explain these concepts. (James Westby, John Arbash Meinel, Aaron Bentley)
- Added bzr log --limit to report a limited number of revisions. (Kent Gibson, #3659)
- Revert does not try to preserve file contents that were originally produced by reverting to a historical revision. (Aaron Bentley)
- bzr log --short now includes [merge] for revisions which have more than one parent. This is a small improvement to help understanding what changes have occurred (John Arbash Meinel, #83887)
- TreeTransform avoids many renames when contructing large trees, improving speed. 3.25x speedups have been observed for construction of kernel-sized-trees, and checkouts are 1.28x faster. (Aaron Bentley)
- Commit on large trees is now faster. In my environment, a commit of a small change to the Mozilla tree (55k files) has dropped from 66 seconds to 32 seconds. For a small tree of 600 files, commit of a small change is 33% faster. (Ian Clatworthy)
- New --create-prefix option to bzr init, like for push. (Daniel Watkins, #56322)
BUGFIXES:
- bzr push should only connect to the remote location one time. We have been connecting 3 times because we forget to pass around the Transport object. This adds BzrDir.clone_on_transport(), so that we can pass in the Transport that we already have. (John Arbash Meinel, #75721)
- DirState.set_state_from_inventory() needs to properly order based on split paths, not just string paths. (John Arbash Meinel, #115947)
- Let TestUIFactoy encode the password prompt with its own stdout. (Vincent Ladeuil, #110204)
- pycurl should take use the range header that takes the range hint into account. (Vincent Ladeuil, #112719)
- WorkingTree4.get_file_sha1 no longer raises an exception when invoked on a missing file. (Aaron Bentley, #118186)
- WorkingTree.remove works correctly with tree references, and when pwd is not the tree root. (Aaron Bentley)
- Merge no longer fails when a file is renamed in one tree and deleted in the other. (Aaron Bentley, #110279)
- revision-info now accepts dotted revnos, doesn't require a tree, and defaults to the last revision (Matthew Fuller, #90048)
- Tests no longer fail when BZR_REMOTE_PATH is set in the environment. (Daniel Watkins, #111958)
- bzr branch -r revid:foo can be used to branch any revision in your repository. (Previously Branch6 only supported revisions in your mainline). (John Arbash Meinel, #115343)
BUGFIXES:
- Handle when you have 2 directories with similar names, but one has a hyphen. ('abc' versus 'abc-2'). The WT4._iter_changes iterator was using direct comparison and 'abc/a' sorts after 'abc-2', but ('abc', 'a') sorts before ('abc-2',). (John Arbash Meinel, #111227)
- Handle when someone renames a file on disk without telling bzr. Previously we would report the first file as missing, but not show the new unknown file. (John Arbash Meinel, #111288)
- Avoid error when running hooks after pulling into or pushing from a branch bound to a smartserver branch. (Martin Pool, #111968)
IMPROVEMENTS:
- Move developer documentation to doc/developers/. This reduces clutter in the root of the source tree and allows HACKING to be split into multiple files. (Robert Collins, Alexander Belchenko)
- Clean up the WorkingTree4._iter_changes() internal loops as well as DirState.update_entry(). This optimizes the core logic for bzr diff and bzr status significantly improving the speed of both. (John Arbash Meinel)
BUGFIXES:
- Handle the case when you delete a file, and then rename another file on top of it. Also handle the case of bzr rm --keep foo. bzr status should show the removed file and an unknown file in its place. (John Arbash Meinel, #109993)
- Bundles properly read and write revision properties that have an empty value. And when the value is not ASCII. (John Arbash Meinel, #109613)
- Fix the bzr commit message to be in text mode. (Alexander Belchenko, #110901)
- Also handle when you rename a file and create a file where it used to be. (John Arbash Meinel, #110256)
- WorkingTree4._iter_changes should not descend into unversioned directories. (John Arbash Meinel, #110399)
NOTES WHEN UPGRADING:
- bzr remove and bzr rm will now remove the working file, if it could be recovered again. This has been done for consistency with svn and the unix rm command. The old remove behaviour has been retained in the new option bzr remove --keep, which will just stop versioning the file, but not delete it. bzr remove --force have been added which will always delete the files. bzr remove is also more verbose. (Marius Kruger, #82602)
IMPROVEMENTS:
- Merge directives can now be supplied as input to merge and pull, like bundles can. (Aaron Bentley)
- Sending the SIGQUIT signal to bzr, which can be done on Unix by pressing Control-Backslash, drops bzr into a debugger. Type 'c' to continue. This can be disabled by setting the environment variable BZR_SIGQUIT_PDB=0. (Martin Pool)
- selftest now supports --list-only to list tests instead of running them. (Ian Clatworthy)
- selftest now supports --exclude PATTERN (or -x PATTERN) to exclude tests with names that match that regular expression. (Ian Clatworthy, #102679)
- selftest now supports --randomize SEED to run tests in a random order. SEED is typically the value 'now' meaning 'use the current time'. (Ian Clatworthy, #102686)
- New option --fixes to commit, which stores bug fixing annotations as revision properties. Built-in support for Launchpad, Debian, Trac and Bugzilla bug trackers. (Jonathan Lange, James Henstridge, Robert Collins)
- New API, bzrlib.bugtracker.tracker_registry, for adding support for other bug trackers to fixes. (Jonathan Lange, James Henstridge, Robert Collins)
- selftest has new short options -f and -1. (Martin Pool)
- bzrlib.tsort.MergeSorter optimizations. Change the inner loop into using local variables instead of going through self._var. Improves the time to merge_sort a 10k revision graph by approximately 40% (~700->400ms). (John Arbash Meinel)
- make docs now creates a man page at man1/bzr.1 fixing bug 107388. (Robert Collins)
- bzr help now provides cross references to other help topics using the _see_also facility on command classes. Likewise the bzr_man documentation, and the bzr.1 man page also include this information. (Robert Collins)
- Tags are now included in logs, that use the long log formatter. (Erik Bågfors, Alexander Belchenko)
- bzr help provides a clearer message when a help topic cannot be found. (Robert Collins, #107656)
- bzr help now accepts optional prefixes for command help. The help for all commands can now be found at bzr help commands/COMMANDNAME as well as bzr help COMMANDNAME (which only works for commands where the name is not the same as a more general help topic). (Robert Collins)
- bzr help PLUGINNAME will now return the module docstring from the plugin PLUGINNAME. (Robert Collins, #50408)
- New help topic urlspec which lists the availables transports. (Goffredo Baroncelli)
- doc/server.txt updated to document the default bzr:// port and also update the blurb about the hpss' current status. (Robert Collins, #107125).
- bzr serve now listens on interface 0.0.0.0 by default, making it serve out to the local LAN (and anyone in the world that can reach the machine running bzr serve. (Robert Collins, #98918)
- A new smart server protocol version has been added. It prefixes requests and responses with an explicit version identifier so that future protocol revisions can be dealt with gracefully. (Andrew Bennetts, Robert Collins)
- The bzr protocol version 2 indicates success or failure in every response without depending on particular commands encoding that consistently, allowing future client refactorings to be much more robust about error handling. (Robert Collins, Martin Pool, Andrew Bennetts)
- The smart protocol over HTTP client has been changed to always post to the same .bzr/smart URL under the original location when it can. This allows HTTP servers to only have to pass URLs ending in .bzr/smart to the smart server handler, and not arbitrary .bzr/*/smart URLs. (Andrew Bennetts)
- digest authentication is now supported for proxies and HTTP by the urllib based http implementation. Tested against Apache 2.0.55 and Squid 2.6.5. Basic and digest authentication are handled coherently for HTTP and proxy: if the user is provided in the url (bzr command line for HTTP, proxy environment variables for proxies), the password is prompted for (only once). If the password is provided, it is taken into account. Once the first authentication is successful, all further authentication roundtrips are avoided by preventively setting the right authentication header(s). (Vincent Ladeuil).
INTERNALS:
- bzrlib API compatability with 0.8 has been dropped, cleaning up some code paths. (Robert Collins)
- Change the format of chroot urls so that they can be safely manipulated by generic url utilities without causing the resulting urls to have escaped the chroot. A side effect of this is that creating a chroot requires an explicit action using a ChrootServer. (Robert Collins, Andrew Bennetts)
- Deprecate Branch.get_root_id() because branches don't have root ids, rather than fixing bug #96847. (Aaron Bentley)
- WorkingTree.apply_inventory_delta provides a better alternative to WorkingTree._write_inventory. (Aaron Bentley)
- Convenience method TestCase.expectFailure ensures that known failures do not silently pass. (Aaron Bentley)
- Transport.local_abspath now raises NotLocalUrl rather than TransportNotPossible. (Martin Pool, Ian Clatworthy)
- New SmartServer hooks facility. There are two initial hooks documented in bzrlib.transport.smart.SmartServerHooks. The two initial hooks allow plugins to execute code upon server startup and shutdown. (Robert Collins).
- SmartServer in standalone mode will now close its listening socket when it stops, rather than waiting for garbage collection. This primarily fixes test suite hangs when a test tries to connect to a shutdown server. It may also help improve behaviour when dealing with a server running on a specific port (rather than dynamically assigned ports). (Robert Collins)
- Move most SmartServer code into a new package, bzrlib/smart. bzrlib/transport/remote.py contains just the Transport classes that used to be in bzrlib/transport/smart.py. (Andrew Bennetts)
- urllib http implementation avoid roundtrips associated with 401 (and 407) errors once the authentication succeeds. (Vincent Ladeuil).
- urlib http now supports querying the user for a proxy password if needed. Realm is shown in the prompt for both HTTP and proxy authentication when the user is required to type a password. (Vincent Ladeuil).
- Renamed SmartTransport (and subclasses like SmartTCPTransport) to RemoteTransport (and subclasses to RemoteTCPTransport, etc). This is more consistent with its new home in bzrlib/transport/remote.py, and because it's not really a "smart" transport, just one that does file operations via remote procedure calls. (Andrew Bennetts)
- The lock_write method of LockableFiles, Repository and Branch now accept a token keyword argument, so that separate instances of those objects can share a lock if it has the right token. (Andrew Bennetts, Robert Collins)
- New method get_branch_reference on BzrDir allows the detection of branch references - which the smart server component needs.
- The Repository API make_working_trees is now permitted to return False when set_make_working_trees is not implemented - previously an unimplemented set_make_working_trees implied the result True from make_working_trees. This has been changed to accomodate the smart server, where it does not make sense (at this point) to ever make working trees by default. (Robert Collins)
- Command objects can now declare related help topics by having _see_also set to a list of related topic. (Robert Collins)
- bzrlib.help now delegates to the Command class for Command specific help. (Robert Collins)
- New class TransportListRegistry, derived from the Registry class, which simplifies tracking the available Transports. (Goffredo Baroncelli)
- New function Branch.get_revision_id_to_revno_map which will return a dictionary mapping revision ids to dotted revnos. Since dotted revnos are defined in the context of the branch tip, it makes sense to generate them from a Branch object. (John Arbash Meinel)
- Fix the 'Unprintable error' message display to use the repr of the exception that prevented printing the error because the str value for it is often not useful in debugging (e.g. KeyError('foo') has a str() of 'foo' but a repr of 'KeyError('foo')' which is much more useful. (Robert Collins)
- urlutils.normalize_url now unescapes unreserved characters, such as "~". (Andrew Bennetts)
BUGFIXES:
- Don't fail bundle selftest if email has 'two' embedded. (Ian Clatworthy, #98510)
- Remove --verbose from bzr bundle. It didn't work anyway. (Robert Widhopf-Fenk, #98591)
- Remove --basis from the checkout/branch commands - it didn't work properly and is no longer beneficial. (Robert Collins, #53675, #43486)
- Don't produce encoding error when adding duplicate files. (Aaron Bentley)
- Fix bzr log <file> so it only logs the revisions that changed the file, and does it faster. (Kent Gibson, John Arbash Meinel, #51980, #69477)
- Fix InterDirstateTre._iter_changes to handle when we come across an empty versioned directory, which now has files in it. (John Arbash Meinel, #104257)
- Teach common_ancestor to shortcut when the tip of one branch is inside the ancestry of the other. Saves a lot of graph processing (with an ancestry of 16k revisions, bzr merge ../already-merged changes from 2m10s to 13s). (John Arbash Meinel, #103757)
- Fix show_diff_trees to handle the case when a file is modified, and the containing directory is renamed. (The file path is different in this versus base, but it isn't marked as a rename). (John Arbash Meinel, #103870)
- FTP now works even when the FTP server does not support atomic rename. (Aaron Bentley, #89436)
- Correct handling in bundles and merge directives of timezones with that are not an integer number of hours offset from UTC. Always represent the epoch time in UTC to avoid problems with formatting earlier times on win32. (Martin Pool, Alexander Belchenko, John Arbash Meinel)
- Typo in the help for register-branch fixed. (Robert Collins, #96770)
- "dirstate" and "dirstate-tags" formats now produce branches compatible with old versions of bzr. (Aaron Bentley, #107168))
- Handle moving a directory when children have been added, removed, and renamed. (John Arbash Meinel, #105479)
- Don't preventively use basic authentication for proxy before receiving a 407 error. Otherwise people willing to use other authentication schemes may expose their password in the clear (or nearly). This add one roundtrip in case basic authentication should be used, but plug the security hole. (Vincent Ladeuil)
- Handle http and proxy digest authentication. (Vincent Ladeuil, #94034).
TESTING:
- Added bzrlib.strace.strace which will strace a single callable and return a StraceResult object which contains just the syscalls involved in running it. (Robert Collins)
- New test method reduceLockdirTimeout to drop the default (ui-centric) default time down to one suitable for tests. (Andrew Bennetts)
- Add new vfs_transport_factory attribute on tests which provides the common vfs backing for both the readonly and readwrite transports. This allows the RemoteObject tests to back onto local disk or memory, and use the existing transport_server attribute all tests know about to be the smart server transport. This in turn allows tests to differentiate between 'transport to access the branch', and 'transport which is a VFS' - which matters in Remote* tests. (Robert Collins, Andrew Bennetts)
- The make_branch_and_tree method for tests will now create a lightweight checkout for the tree if the vfs_transport_factory is not a LocalURLServer. (Robert Collins, Andrew Bennetts)
- Branch implementation tests have been audited to ensure that all urls passed to Branch APIs use proper urls, except when local-disk paths are intended. This is so that tests correctly access the test transport which is often not equivalent to local disk in Remote* tests. As part of this many tests were adjusted to remove dependencies on local disk access. (Robert Collins, Andrew Bennetts)
- Mark bzrlib.tests and bzrlib.tests.TestUtil as providing assertFOO helper functions by adding a __unittest global attribute. (Robert Collins, Andrew Bennetts, Martin Pool, Jonathan Lange)
- Refactored proxy and authentication handling to simplify the implementation of new auth schemes for both http and proxy. (Vincent Ladeuil)
BUGFIXES:
- Handle incompatible repositories as a user issue when fetching. (Aaron Bentley)
- Don't give a recommendation to upgrade when branching or checking out a branch that contains an old-format working tree. (Martin Pool)
CHANGES:
- A warning is now displayed when opening working trees in older formats, to encourage people to upgrade to WorkingTreeFormat4. (Martin Pool)
IMPROVEMENTS:
- HTTP redirections are now taken into account when a branch (or a bundle) is accessed for the first time. A message is issued at each redirection to inform the user. In the past, http redirections were silently followed for each request which significantly degraded the performances. The http redirections are not followed anymore by default, instead a RedirectRequested exception is raised. For bzrlib users needing to follow http redirections anyway, bzrlib.transport.do_catching_redirections provide an easy transition path. (vila)
INTERNALS:
- Added ReadLock.temporary_write_lock() to allow upgrading an OS read lock to an OS write lock. Linux can do this without unlocking, Win32 needs to unlock in between. (John Arbash Meinel)
- New parameter recommend_upgrade to BzrDir.open_workingtree to silence (when false) warnings about opening old formats. (Martin Pool)
- Fix minor performance regression with bzr-0.15 on pre-dirstate trees. (We were reading the working inventory too many times). (John Arbash Meinel)
- Remove Branch.get_transaction() in favour of a simple cache of revision_history. Branch subclasses should override _gen_revision_history rather than revision_history to make use of this cache, and call _clear_revision_history_cache and _cache_revision_history at appropriate times. (Andrew Bennetts)
BUGFIXES:
- Take smtp_server from user config into account. (vila, #92195)
- Restore Unicode filename handling for versioned and unversioned files. (John Arbash Meinel, #92608)
- Don't fail during bzr commit if a file is marked removed, and the containing directory is auto-removed. (John Arbash Meinel, #93681)
- bzr status FILENAME failed on Windows because of an uncommon errno. (ERROR_DIRECTORY == 267 != ENOTDIR). (Wouter van Heyst, John Arbash Meinel, #90819)
- bzr checkout source should create a local branch in the same format as source. (John Arbash Meinel, #93854)
- bzr commit with a kind change was failing to update the last-changed-revision for directories. The InventoryDirectory._unchanged only looked at the parent_id and name, ignoring the fact that the kind could have changed, too. (John Arbash Meinel, #90111)
- bzr mv dir/subdir other was incorrectly updating files inside the directory. So that there was a chance it would break commit, etc. (John Arbash Meinel, #94037)
- Correctly handles mutiple permanent http redirections. (vila, #88780)
NOTES WHEN UPGRADING:
- Release 0.15rc2 of bzr changes the bzr init-repo command to default to --trees instead of --no-trees. Existing shared repositories are not affected.
IMPROVEMENTS:
- New merge-directive command to generate machine- and human-readable merge requests. (Aaron Bentley)
- New submit: revision specifier makes it easy to diff against the common ancestor with the submit location (Aaron Bentley)
- Added support for Putty's SSH implementation. (Dmitry Vasiliev)
- Added bzr status --versioned to report only versioned files, not unknowns. (Kent Gibson)
- Merge now autodetects the correct line-ending style for its conflict markers. (Aaron Bentley)
INTERNALS:
- Refactored SSH vendor registration into SSHVendorManager class. (Dmitry Vasiliev)
BUGFIXES:
- New --numbered-dirs option to bzr selftest to use numbered dirs for TestCaseInTempDir. This is default behavior on Windows. Anyone can force named dirs on Windows with --no-numbered-dirs. (Alexander Belchenko)
- Fix RevisionSpec_revid to handle the Unicode strings passed in from the command line. (Marien Zwart, #90501)
- Fix TreeTransform._iter_changes when both the source and destination are missing. (Aaron Bentley, #88842)
- Fix commit of merges with symlinks in dirstate trees. (Marien Zwart)
- Switch the bzr init-repo default from --no-trees to --trees. (Wouter van Heyst, #53483)
SURPRISES:
- The default disk format has changed. Please run 'bzr upgrade' in your working trees to upgrade. This new default is compatible for network operations, but not for local operations. That is, if you have two versions of bzr installed locally, after upgrading you can only use the bzr 0.15 version. This new default does not enable tags or nested-trees as they are incompatible with bzr versions before 0.15 over the network.
- For users of bzrlib: Two major changes have been made to the working tree api in bzrlib. The first is that many methods and attributes, including the inventory attribute, are no longer valid for use until one of lock_read/lock_write/lock_tree_write has been called, and become invalid again after unlock is called. This has been done to improve performance and correctness as part of the dirstate development. (Robert Collins, John A Meinel, Martin Pool, and others).
- For users of bzrlib: The attribute 'tree.inventory' should be considered readonly. Previously it was possible to directly alter this attribute, or its contents, and have the tree notice this. This has been made unsupported - it may work in some tree formats, but in the newer dirstate format such actions will have no effect and will be ignored, or even cause assertions. All operations possible can still be carried out by a combination of the tree API, and the bzrlib.transform API. (Robert Collins, John A Meinel, Martin Pool, and others).
IMPROVEMENTS:
- Support for OS Windows 98. Also .bzr.log on any windows system saved in My Documents folder. (Alexander Belchenko)
- bzr mv enhanced to support already moved files. In the past the mv command would have failed if the source file doesn't exist. In this situation bzr mv would now detect that the file has already moved and update the repository accordingly, if the target file does exist. A new option --after has been added so that if two files already exist, you could notify Bazaar that you have moved a (versioned) file and replaced it with another. Thus in this case bzr move --after will only update the Bazaar identifier. (Steffen Eichenberg, Marius Kruger)
- ls now works on treeless branches and remote branches. (Aaron Bentley)
- bzr help global-options describes the global options. (Aaron Bentley)
- bzr pull --overwrite will now correctly overwrite checkouts. (Robert Collins)
- Files are now allowed to change kind (e.g. from file to symlink). Supported by commit, revert and status (Aaron Bentley)
- inventory and unknowns hidden in favour of ls (Aaron Bentley)
- bzr help checkouts descibes what checkouts are and some possible uses of them. (James Westby, Aaron Bentley)
- A new -d option to push, pull and merge overrides the default directory. (Martin Pool)
- Branch format 6: smaller, and potentially faster than format 5. Supports append_history_only mode, where the log view and revnos do not change, except by being added to. Stores policy settings in ".bzr/branch/branch.conf".
- append_only branches: Format 6 branches may be configured so that log view and revnos are always consistent. Either create the branch using "bzr init --append-revisions-only" or edit the config file as descriped in docs/configuration.txt.
- rebind: Format 6 branches retain the last-used bind location, so if you "bzr unbind", you can "bzr bind" to bind to the previously-selected bind location.
- Builtin tags support, created and deleted by the tag command and stored in the branch. Tags can be accessed with the revisionspec -rtag:, and listed with bzr tags. Tags are not versioned at present. Tags require a network incompatible upgrade. To perform this upgrade, run bzr upgrade --dirstate-tags in your branch and repositories. (Martin Pool)
- The bzr:// transport now has a well-known port number, 4155, which it will use by default. (Andrew Bennetts, Martin Pool)
- Bazaar now looks for user-installed plugins before looking for site-wide plugins. (Jonathan Lange)
- bzr resolve now detects and marks resolved text conflicts. (Aaron Bentley)
INTERNALS:
- Internally revision ids and file ids are now passed around as utf-8 bytestrings, rather than treating them as Unicode strings. This has performance benefits for Knits, since we no longer need to decode the revision id for each line of content, nor for each entry in the index. This will also help with the future dirstate format. (John Arbash Meinel)
- Reserved ids (any revision-id ending in a colon) are rejected by versionedfiles, repositories, branches, and working trees (Aaron Bentley)
- Minor performance improvement by not creating a ProgressBar for every KnitIndex we create. (about 90ms for a bzr.dev tree) (John Arbash Meinel)
- New easier to use Branch hooks facility. There are five initial hooks, all documented in bzrlib.branch.BranchHooks.__init__ - 'set_rh', 'post_push', 'post_pull', 'post_commit', 'post_uncommit'. These hooks fire after the matching operation on a branch has taken place, and were originally added for the branchrss plugin. (Robert Collins)
- New method Branch.push() which should be used when pushing from a branch as it makes performance and policy decisions to match the UI level command push. (Robert Collins).
- Add a new method Tree.revision_tree which allows access to cached trees for arbitrary revisions. This allows the in development dirstate tree format to provide access to the callers to cached copies of inventory data which are cheaper to access than inventories from the repository. (Robert Collins, Martin Pool)
- New Branch.last_revision_info method, this is being done to allow optimization of requests for both the number of revisions and the last revision of a branch with smartservers and potentially future branch formats. (Wouter van Heyst, Robert Collins)
- Allow 'import bzrlib.plugins.NAME' to work when the plugin NAME has not yet been loaded by load_plugins(). This allows plugins to depend on each other for code reuse without requiring users to perform file-renaming gymnastics. (Robert Collins)
- New Repository method 'gather_stats' for statistic data collection. This is expected to grow to cover a number of related uses mainly related to bzr info. (Robert Collins)
- Log formatters are now managed with a registry. log.register_formatter continues to work, but callers accessing the FORMATTERS dictionary directly will not.
- Allow a start message to be passed to the edit_commit_message function. This will be placed in the message offered to the user for editing above the separator. It allows a template commit message to be used more easily. (James Westby)
- GPGStrategy.sign() will now raise BzrBadParameterUnicode if you pass a Unicode string rather than an 8-bit string. Callers need to be updated to encode first. (John Arbash Meinel)
- Branch.push, pull, merge now return Result objects with information about what happened, rather than a scattering of various methods. These are also passed to the post hooks. (Martin Pool)
- File formats and architecture is in place for managing a forest of trees in bzr, and splitting up existing trees into smaller subtrees, and finally joining trees to make a larger tree. This is the first iteration of this support, and the user-facing aspects still require substantial work. If you wish to experiment with it, use bzr upgrade --dirstate-with-subtree in your working trees and repositories. You can use the hidden commands split and join and to create and manipulate nested trees, but please consider using the nested-trees branch, which contains substantial UI improvements, instead. http://code.aaronbentley.com/bzr/bzrrepo/nested-trees/ (Aaron Bentley, Martin Pool, Robert Collins).
BUGFIXES:
- bzr annotate now uses dotted revnos from the viewpoint of the branch, rather than the last changed revision of the file. (John Arbash Meinel, #82158)
- Lock operations no longer hang if they encounter a permission problem. (Aaron Bentley)
- bzr push can resume a push that was canceled before it finished. Also, it can push even if the target directory exists if you supply the --use-existing-dir flag. (John Arbash Meinel, #30576, #45504)
- Fix http proxy authentication when user and an optional password appears in the *_proxy vars. (Vincent Ladeuil, #83954).
- bzr log branch/file works for local treeless branches (Aaron Bentley, #84247)
- Fix problem with UNC paths on Windows 98. (Alexander Belchenko, #84728)
- Searching location of CA bundle for PyCurl in env variable (CURL_CA_BUNDLE), and on win32 along the PATH. (Alexander Belchenko, #82086)
- bzr init works with unicode argument LOCATION. (Alexander Belchenko, #85599)
- Raise DependencyNotPresent if pycurl do not support https. (Vincent Ladeuil, #85305)
- Invalid proxy env variables should not cause a traceback. (Vincent Ladeuil, #87765)
- Ignore patterns normalised to use '/' path separator. (Kent Gibson, #86451)
- bzr rocks. It sure does! Fix case. (Vincent Ladeuil, #78026)
- Fix bzrtools shelve command for removed lines beginning with "--" (Johan Dahlberg, #75577)
TESTING:
- New --first option to bzr selftest to run specified tests before the rest of the suite. (Martin Pool)
IMPROVEMENTS:
- bzr help global-options describes the global options. (Aaron Bentley)
BUG FIXES:
- Skip documentation generation tests if the tools to do so are not available. Fixes running selftest for installled copies of bzr. (John Arbash Meinel, #80330)
- Fix the code that discovers whether bzr is being run from it's working tree to handle the case when it isn't but the directory it is in is below a repository. (James Westby, #77306)
IMPROVEMENTS:
New connection: bzr+http:// which supports tunnelling the smart protocol over an HTTP connection. If writing is enabled on the bzr server, then you can write over the http connection. (Andrew Bennetts, John Arbash Meinel)
Aliases now support quotation marks, so they can contain whitespace (Marius Kruger)
PyCurlTransport now use a single curl object. By specifying explicitly the 'Range' header, we avoid the need to use two different curl objects (and two connections to the same server). (Vincent Ladeuil)
bzr commit does not prompt for a message until it is very likely to succeed. (Aaron Bentley)
bzr conflicts now takes --text to list pathnames of text conflicts (Aaron Bentley)
Fix iter_lines_added_or_present_in_versions to use a set instead of a list while checking if a revision id was requested. Takes 10s off of the fileids_affected_by_revision_ids time, which is 10s of the bzr branch time. Also improve fileids_... time by filtering lines with a regex rather than multiple str.find() calls. (saves another 300ms) (John Arbash Meinel)
Policy can be set for each configuration key. This allows keys to be inherited properly across configuration entries. For example, this should enable you to do:
[/home/user/project] push_location = sftp://host/srv/project/ push_location:policy = appendpathAnd then a branch like /home/user/project/mybranch should get an automatic push location of sftp://host/srv/project/mybranch. (James Henstridge)
Added bzr status --short to make status report svn style flags for each file. For example:
$ bzr status --short A foo A bar D baz ? wooley'bzr selftest --clean-output' allows easily clean temporary tests directories without running tests. (Alexander Belchenko)
bzr help hidden-commands lists all hidden commands. (Aaron Bentley)
bzr merge now has an option --pull to fall back to pull if local is fully merged into remote. (Jan Hudec)
bzr help formats describes available directory formats. (Aaron Bentley)
INTERNALS:
- A few tweaks directly to fileids_affected_by_revision_ids to help speed up processing, as well allowing to extract unannotated lines. Between the two fileids_affected_by_revision_ids is improved by approx 10%. (John Arbash Meinel)
- Change Revision serialization to only write out millisecond resolution. Rather than expecting floating point serialization to preserve more resolution than we need. (Henri Weichers, Martin Pool)
- Test suite ends cleanly on Windows. (Vincent Ladeuil)
- When encoding_type attribute of class Command is equal to 'exact', force sys.stdout to be a binary stream on Windows, and therefore keep exact line-endings (without LF -> CRLF conversion). (Alexander Belchenko)
- Single-letter short options are no longer globally declared. (Martin Pool)
- Before using detected user/terminal encoding bzr should check that Python has corresponding codec. (Alexander Belchenko)
- Formats for end-user selection are provided via a FormatRegistry (Aaron Bentley)
BUG FIXES:
- bzr missing --verbose was showing adds/removals in the wrong direction. (John Arbash Meinel)
- bzr annotate now defaults to showing dotted revnos for merged revisions. It cuts them off at a depth of 12 characters, but you can supply --long to see the full number. You can also use --show-ids to display the original revision ids, rather than revision numbers and committer names. (John Arbash Meinel, #75637)
- bzr now supports Win32 UNC path (e.g. \HOST\path. (Alexander Belchenko, #57869)
- Win32-specific: output of cat, bundle and diff commands don't mangle line-endings (Alexander Belchenko, #55276)
- Replace broken fnmatch based ignore pattern matching with custom pattern matcher. (Kent Gibson, Jan Hudec #57637)
- pycurl and urllib can detect short reads at different places. Update the test suite to test more cases. Also detect http error code 416 which was raised for that specific bug. Also enhance the urllib robustness by detecting invalid ranges (and pycurl's one by detecting short reads during the initial GET). (Vincent Ladeuil, #73948)
- The urllib connection sharing interacts badly with urllib2 proxy setting (the connections didn't go thru the proxy anymore). Defining a proper ProxyHandler solves the problem. (Vincent Ladeuil, #74759)
- Use urlutils to generate relative URLs, not osutils (Aaron Bentley, #76229)
- bzr status in a readonly directory should work without giving lots of errors. (John Arbash Meinel, #76299)
- Mention the revisionspec topic for the revision option help. (Wouter van Heyst, #31663)
- Allow plugins import from zip archives. (Alexander Belchenko, #68124)
No changes from 0.13rc1
IMPROVEMENTS:
- New command bzr remove-tree allows the removal of the working tree from a branch. (Daniel Silverstone)
- urllib uses shared keep-alive connections, so http operations are substantially faster. (Vincent Ladeuil, #53654)
- bzr export allows an optional branch parameter, to export a bzr tree from some other url. For example: bzr export bzr.tar.gz http://bazaar-vcs.org/bzr/bzr.dev (Daniel Silverstone)
- Added bzr help topics to the bzr help system. This gives a location for general information, outside of a specific command. This includes updates for bzr help revisionspec the first topic included. (Goffredo Baroncelli, John Arbash Meinel, #42714)
- WSGI-compatible HTTP smart server. See doc/http_smart_server.txt. (Andrew Bennetts)
- Knit files will now cache full texts only when the size of the deltas is as large as the size of the fulltext. (Or after 200 deltas, whichever comes first). This has the most benefit on large files with small changes, such as the inventory for a large project. (eg For a project with 2500 files, and 7500 revisions, it changes the size of inventory.knit from 11MB to 5.4MB) (John Arbash Meinel)
INTERNALS:
- New -D option given before the command line turns on debugging output for particular areas. -Derror shows tracebacks on all errors. (Martin Pool)
- Clean up bzr selftest --benchmark bundle to correct an import, and remove benchmarks that take longer than 10min to run. (John Arbash Meinel)
- Use time.time() instead of time.clock() to decide on progress throttling. Because time.clock() is actually CPU time, so over a high-latency connection, too many updates get throttled. (John Arbash Meinel)
- MemoryTransport.list_dir() would strip the first character for files or directories in root directory. (John Arbash Meinel)
- New method get_branch_reference on 'BzrDir' allows the detection of branch references - which the smart server component needs.
- New ChrootTransportDecorator, accessible via the chroot+ url prefix. It disallows any access to locations above a set URL. (Andrew Bennetts)
BUG FIXES:
- Now _KnitIndex properly decode revision ids when loading index data. And optimize the knit index parsing code. (Dmitry Vasiliev, John Arbash Meinel)
- bzrlib/bzrdir.py was directly referencing bzrlib.workingtree, without importing it. This prevented bzr upgrade from working unless a plugin already imported bzrlib.workingtree (John Arbash Meinel, #70716)
- Suppress the traceback on invalid URLs (Vincent Ladeuil, #70803).
- Give nicer error message when an http server returns a 403 error code. (Vincent Ladeuil, #57644).
- When a multi-range http GET request fails, try a single range one. If it fails too, forget about ranges. Remember that until the death of the transport and propagates that to the clones. (Vincent Ladeuil, #62276, #62029).
- Handles user/passwords supplied in url from command line (for the urllib implementation). Don't request already known passwords (Vincent Ladeuil, #42383, #44647, #48527)
- _KnitIndex.add_versions() dictionary compresses revision ids as they are added. This fixes bug where fetching remote revisions records them as full references rather than integers. (John Arbash Meinel, #64789)
- bzr ignore strips trailing slashes in patterns. Also bzr ignore rejects absolute paths. (Kent Gibson, #4559)
- bzr ignore takes multiple arguments. (Cheuksan Edward Wang, #29488)
- mv correctly handles paths that traverse symlinks. (Aaron Bentley, #66964)
- Give nicer looking error messages when failing to connect over ssh. (John Arbash Meinel, #49172)
- Pushing to a remote branch does not currently update the remote working tree. After a remote push, bzr status and bzr diff on the remote machine now show that the working tree is out of date. (Cheuksan Edward Wang #48136)
- Use patiencediff instead of difflib for determining deltas to insert into knits. This avoids the O(N^3) behavior of difflib. Patience diff should be O(N^2). (Cheuksan Edward Wang, #65714)
- Running bzr log on nonexistent file gives an error instead of the entire log history. (Cheuksan Edward Wang #50793)
- bzr cat can look up contents of removed or renamed files. If the pathname is ambiguous, i.e. the files in the old and new trees have different id's, the default is the file in the new tree. The user can use "--name-from-revision" to select the file in the old tree. (Cheuksan Edward Wang, #30190)
TESTING:
- TestingHTTPRequestHandler really handles the Range header (previously it was ignoring it and returning the whole file,).
INTERNALS:
- Clean up bzr selftest --benchmark bundle to correct an import, and remove benchmarks that take longer than 10min to run. (John Arbash Meinel)
IMPROVEMENTS:
- bzr log now shows dotted-decimal revision numbers for all revisions, rather than just showing a decimal revision number for revisions on the mainline. These revision numbers are not yet accepted as input into bzr commands such as log, diff etc. (Robert Collins)
- revisions can now be specified using dotted-decimal revision numbers. For instance, bzr diff -r 1.2.1..1.2.3. (Robert Collins)
- bzr help commands output is now shorter (Aaron Bentley)
- bzr now uses lazy importing to reduce the startup time. This has a moderate effect on lots of actions, especially ones that have little to do. For example bzr rocks time is down to 116ms from 283ms. (John Arbash Meinel)
- New Registry class to provide name-to-object registry-like support, for example for schemes where plugins can register new classes to do certain tasks (e.g. log formatters). Also provides lazy registration to allow modules to be loaded on request. (John Arbash Meinel, Adeodato Simó)
API INCOMPATABILITY:
- LogFormatter subclasses show now expect the 'revno' parameter to show() to be a string rather than an int. (Robert Collins)
INTERNALS:
- TestCase.run_bzr, run_bzr_captured, and run_bzr_subprocess can take a working_dir='foo' parameter, which will change directory for the command. (John Arbash Meinel)
- bzrlib.lazy_regex.lazy_compile can be used to create a proxy around a regex, which defers compilation until first use. (John Arbash Meinel)
- TestCase.run_bzr_subprocess defaults to supplying the --no-plugins parameter to ensure test reproducability, and avoid problems with system-wide installed plugins. (John Arbash Meinel)
- Unique tree root ids are now supported. Newly created trees still use the common root id for compatibility with bzr versions before 0.12. (Aaron Bentley)
- WorkingTree.set_root_id(None) is now deprecated. Please pass in inventory.ROOT_ID if you want the default root id value. (Robert Collins, John Arbash Meinel)
- New method WorkingTree.flush() which will write the current memory inventory out to disk. At the same time, read_working_inventory will no longer trash the current tree inventory if it has been modified within the current lock, and the tree will now flush() automatically on unlock(). WorkingTree.set_root_id() has been updated to take advantage of this functionality. (Robert Collins, John Arbash Meinel)
- bzrlib.tsort.merge_sorted now accepts generate_revnos. This parameter will cause it to add another column to its output, which contains the dotted-decimal revno for each revision, as a tuple. (Robert Collins)
- LogFormatter.show_merge is deprecated in favour of LogFormatter.show_merge_revno. (Robert Collins)
BUG FIXES:
- Avoid circular imports by creating a deprecated function for bzrlib.tree.RevisionTree. Callers should have been using bzrlib.revisontree.RevisionTree anyway. (John Arbash Meinel, #63360, #66349)
- Don't use socket.MSG_WAITALL as it doesn't exist on all platforms. (Martin Pool, #66356)
- Don't require Content-Type in range responses. Assume they are a single range if Content-Type does not exist. (John Arbash Meinel, #62473)
- bzr branch/pull no longer complain about progress bar cleanup when interrupted during fetch. (Aaron Bentley, #54000)
- WorkingTree.set_parent_trees() uses the trees to directly write the basis inventory, rather than going through the repository. This allows us to have 1 inventory read, and 2 inventory writes when committing a new tree. (John Arbash Meinel)
- When reverting, files that are not locally modified that do not exist in the target are deleted, not just unversioned (Aaron Bentley)
- When trying to acquire a lock, don't fail immediately. Instead, try a few times (up to 1 hour) before timing out. Also, report why the lock is unavailable (John Arbash Meinel, #43521, #49556)
- Leave HttpTransportBase daughter classes decides how they implement cloning. (Vincent Ladeuil, #61606)
- diff3 does not indicate conflicts on clean merge. (Aaron Bentley)
- If a commit fails, the commit message is stored in a file at the root of the tree for later commit. (Cheuksan Edward Wang, Stefan Metzmacher, #32054)
TESTING:
- New test base class TestCaseWithMemoryTransport offers memory-only testing facilities: its not suitable for tests that need to mutate disk state, but most tests should not need that and should be converted to TestCaseWithMemoryTransport. (Robert Collins)
- TestCase.make_branch_and_memory_tree now takes a format option to set the BzrDir, Repository and Branch formats of the created objects. (Robert Collins, John Arbash Meinel)
- Smart server transport test failures on windows fixed. (Lukáš Lalinský).
BUG FIXES:
- Test suite hangs on windows fixed. (Andrew Bennets, Alexander Belchenko).
- Commit performance regression fixed. (Aaron Bentley, Robert Collins, John Arbash Meinel).
IMPROVEMENTS:
- Knit files now wait to create their contents until the first data is added. The old code used to create an empty .knit and a .kndx with just the header. However, this caused a lot of extra round trips over sftp. This can change the time for bzr push to create a new remote branch from 160s down to 100s. This also affects bzr commit performance when adding new files, bzr commit on a new kernel-like tree drops from 50s down to 40s (John Arbash Meinel, #44692)
- When an entire subtree has been deleted, commit will now report that just the top of the subtree has been deleted, rather than reporting all the individual items. (Robert Collins)
- Commit performs one less XML parse. (Robert Collins)
- bzr checkout now operates on readonly branches as well as readwrite branches. This fixes bug #39542. (Robert Collins)
- bzr bind no longer synchronises history with the master branch. Binding should be followed by an update or push to synchronise the two branches. This is closely related to the fix for bug #39542. (Robert Collins)
- bzrlib.lazy_import.lazy_import function to create on-demand objects. This allows all imports to stay at the global scope, but modules will not actually be imported if they are not used. (John Arbash Meinel)
- Support bzr:// and bzr+ssh:// urls to work with the new RPC-based transport which will be used with the upcoming high-performance smart server. The new command bzr serve will invoke bzr in server mode, which processes these requests. (Andrew Bennetts, Robert Collins, Martin Pool)
- New command bzr version-info which can be used to get a summary of the current state of the tree. This is especially useful as part of a build commands. See doc/version_info.txt for more information (John Arbash Meinel)
BUG FIXES:
- 'bzr inventory [FILE...]' allows restricting the file list to a specific set of files. (John Arbash Meinel, #3631)
- Don't abort when annotating empty files (John Arbash Meinel, #56814)
- Add Stanza.to_unicode() which can be passed to another Stanza when nesting stanzas. Also, add read_stanza_unicode to handle when reading a nested Stanza. (John Arbash Meinel)
- Transform._set_mode() needs to stat the right file. (John Arbash Meinel, #56549)
- Raise WeaveFormatError rather than StopIteration when trying to read an empty Weave file. (John Arbash Meinel, #46871)
- Don't access e.code for generic URLErrors, only HTTPErrors have .code. (Vincent Ladeuil, #59835)
- Handle boundary="" lines properly to allow access through a Squid proxy. (John Arbash Meinel, #57723)
- revert now removes newly-added directories (Aaron Bentley, #54172)
- bzr upgrade sftp:// shouldn't fail to upgrade v6 branches if there isn't a working tree. (David Allouche, #40679)
- Give nicer error messages when a user supplies an invalid --revision parameter. (John Arbash Meinel, #55420)
- Handle when LANG is not recognized by python. Emit a warning, but just revert to using 'ascii'. (John Arbash Meinel, #35392)
- Don't use preexec_fn on win32, as it is not supported by subprocess. (John Arbash Meinel)
- Skip specific tests when the dependencies aren't met. This includes some setup.py tests when python-dev is not available, and some tests that depend on paramiko. (John Arbash Meinel, Mattheiu Moy)
- Fallback to Paramiko properly, if no ssh executable exists on the system. (Andrew Bennetts, John Arbash Meinel)
- Branch.bind(other_branch) no longer takes a write lock on the other branch, and will not push or pull between the two branches. API users will need to perform a push or pull or update operation if they require branch synchronisation to take place. (Robert Collins, #47344)
- When creating a tarball or zipfile export, export unicode names as utf-8 paths. This may not work perfectly on all platforms, but has the best chance of working in the common case. (John Arbash Meinel, #56816)
- When committing, only files that exist in working tree or basis tree may be specified (Aaron Bentley, #50793)
PORTABILITY:
- Fixes to run on Python 2.5 (Brian M. Carlson, Martin Pool, Marien Zwart)
INTERNALS:
- TestCaseInTempDir now creates a separate directory for HOME, rather than having HOME set to the same location as the working directory. (John Arbash Meinel)
- run_bzr_subprocess() can take an optional env_changes={} parameter, which will update os.environ inside the spawned child. It also can take a universal_newlines=True, which helps when checking the output of the command. (John Arbash Meinel)
- Refactor SFTP vendors to allow easier re-use when ssh is used. (Andrew Bennetts)
- Transport.list_dir() and Transport.iter_files_recursive() should always return urlescaped paths. This is now tested (there were bugs in a few of the transports) (Andrew Bennetts, David Allouche, John Arbash Meinel)
- New utility function symbol_versioning.deprecation_string. Returns the formatted string for a callable, deprecation format pair. (Robert Collins)
- New TestCase helper applyDeprecated. This allows you to call a callable which is deprecated without it spewing to the screen, just by supplying the deprecation format string issued for it. (Robert Collins)
- Transport.append and Transport.put have been deprecated in favor of .append_bytes, .append_file, .put_bytes, and .put_file. This removes the ambiguity in what type of object the functions take. Transport.non_atomic_put_{bytes,file} has also been added. Which works similarly to Transport.append() except for SFTP, it doesn't have a round trip when opening the file. Also, it provides functionality for creating a parent directory when trying to create a file, rather than raise NoSuchFile and forcing the caller to repeat their request. (John Arbash Meinel)
- WorkingTree has a new api unversion which allow the unversioning of entries by their file id. (Robert Collins)
- WorkingTree.pending_merges is deprecated. Please use the get_parent_ids (introduced in 0.10) method instead. (Robert Collins)
- WorkingTree has a new lock_tree_write method which locks the branch for read rather than write. This is appropriate for actions which only need the branch data for reference rather than mutation. A new decorator needs_tree_write_lock is provided in the workingtree module. Like the needs_read_lock and needs_write_lock decorators this allows static declaration of the locking requirements of a function to ensure that a lock is taken out for casual scripts. (Robert Collins, #54107)
- All WorkingTree methods which write to the tree, but not to the branch have been converted to use needs_tree_write_lock rather than needs_write_lock. Also converted is the revert, conflicts and tree transform modules. This provides a modest performance improvement on metadir style trees, due to the reduce lock-acquisition, and a more significant performance improvement on lightweight checkouts from remote branches, where trivial operations used to pay a significant penalty. It also provides the basis for allowing readonly checkouts. (Robert Collins)
- Special case importing the standard library 'copy' module. This shaves off 40ms of startup time, while retaining compatibility. See: bzrlib/inspect_for_copy.py for more details. (John Arbash Meinel)
- WorkingTree has a new parent class MutableTree which represents the specialisations of Tree which are able to be altered. (Robert Collins)
- New methods mkdir and put_file_bytes_non_atomic on MutableTree that mutate the tree and its contents. (Robert Collins)
- Transport behaviour at the root of the URL is now defined and tested. (Andrew Bennetts, Robert Collins)
TESTING:
- New test helper classs MemoryTree. This is typically accessed via self.make_branch_and_memory_tree() in test cases. (Robert Collins)
- Add start_bzr_subprocess and stop_bzr_subprocess to allow test code to continue running concurrently with a subprocess of bzr. (Andrew Bennetts, Robert Collins)
- Add a new method Transport.get_smart_client(). This is provided to allow upgrades to a richer interface than the VFS one provided by Transport. (Andrew Bennetts, Martin Pool)
- IMPROVEMENTS:
- 'merge' now takes --uncommitted, to apply uncommitted changes from a tree. (Aaron Bentley)
- 'bzr add --file-ids-from' can be used to specify another path to use for creating file ids, rather than generating all new ones. Internally, the 'action' passed to smart_add_tree() can return file_ids that will be used, rather than having bzrlib generate new ones. (John Arbash Meinel, #55781)
- bzr selftest --benchmark now allows a --cache-dir parameter. This will cache some of the intermediate trees, and decrease the setup time for benchmark tests. (John Arbash Meinel)
- Inverse forms are provided for all boolean options. For example, --strict has --no-strict, --no-recurse has --recurse (Aaron Bentley)
- Serialize out Inventories directly, rather than using ElementTree. Writing out a kernel sized inventory drops from 2s down to ~350ms. (Robert Collins, John Arbash Meinel)
BUG FIXES:
- Help diffutils 2.8.4 get along with binary tests (Marien Zwart: #57614)
- Change LockDir so that if the lock directory doesn't exist when lock_write() is called, an attempt will be made to create it. (John Arbash Meinel, #56974)
- bzr uncommit preserves pending merges. (John Arbash Meinel, #57660)
- Active FTP transport now works as intended. (ghozzy, #56472)
- Really fix mutter() so that it won't ever raise a UnicodeError. It means it is possible for ~/.bzr.log to contain non UTF-8 characters. But it is a debugging log, not a real user file. (John Arbash Meinel, #56947, #53880)
- Change Command handle to allow Unicode command and options. At present we cannot register Unicode command names, so we will get BzrCommandError('unknown command'), or BzrCommandError('unknown option') But that is better than a UnicodeError + a traceback. (John Arbash Meinel, #57123)
- Handle TZ=UTC properly when reading/writing revisions. (John Arbash Meinel, #55783, #56290)
- Use GPG_TTY to allow gpg --cl to work with gpg-agent in a pipeline, (passing text to sign in on stdin). (John Arbash Meinel, #54468)
- External diff does the right thing for binaries even in foreign languages. (John Arbash Meinel, #56307)
- Testament handles more cases when content is unicode. Specific bug was in handling of revision properties. (John Arbash Meinel, Holger Krekel, #54723)
- The bzr selftest was failing on installed versions due to a bug in a new test helper. (John Arbash Meinel, Robert Collins, #58057)
INTERNALS:
- bzrlib.cache_utf8 contains encode() and decode() functions which can be used to cache the conversion between utf8 and Unicode. Especially helpful for some of the knit annotation code, which has to convert revision ids to utf8 to annotate lines in storage. (John Arbash Meinel)
- setup.py now searches the filesystem to find all packages which need to be installed. This should help make the life of packagers easier. (John Arbash Meinel)
SURPRISES:
- The hard-coded built-in ignore rules have been removed. There are now two rulesets which are enforced. A user global one in ~/.bazaar/ignore which will apply to every tree, and the tree specific one '.bzrignore'. ~/.bazaar/ignore will be created if it does not exist, but with a more conservative list than the old default. This fixes bugs with default rules being enforced no matter what. The old list of ignore rules from bzr is available by running 'bzr ignore --old-default-rules'. (Robert Collins, Martin Pool, John Arbash Meinel)
- 'branches.conf' has been changed to 'locations.conf', since it can apply to more locations than just branch locations. (Aaron Bentley)
IMPROVEMENTS:
- The revision specifier "revno:" is extended to accept the syntax revno:N:branch. For example, revno:42:http://bazaar-vcs.org/bzr/bzr.dev/ means revision 42 in bzr.dev. (Matthieu Moy)
- Tests updates to ensure proper URL handling, UNICODE support, and proper printing when the user's terminal encoding cannot display the path of a file that has been versioned. bzr branch can take a target URL rather than only a local directory. Branch.get_parent()/set_parent() now save a relative path if possible, and normalize the parent based on root, allowing access across different transports. (John Arbash Meinel, Wouter van Heyst, Martin Pool) (Malone #48906, #42699, #40675, #5281, #3980, #36363, #43689, #42517, #42514)
- On Unix, detect terminal width using an ioctl not just $COLUMNS. Use terminal width for single-line logs from bzr log --line and pending-merge display. (Robert Widhopf-Fenk, Gustavo Niemeyer) (Malone #3507)
- On Windows, detect terminal width using GetConsoleScreenBufferInfo. (Alexander Belchenko)
- Speedup improvement for 'date:'-revision search. (Guillaume Pinot).
- Show the correct number of revisions pushed when pushing a new branch. (Robert Collins).
- 'bzr selftest' now shows a progress bar with the number of tests, and progress made. 'make check' shows tests in -v mode, to be more useful for the PQM status window. (Robert Collins). When using a progress bar, failed tests are printed out, rather than being overwritten by the progress bar until the suite finishes. (John Arbash Meinel)
- 'bzr selftest --benchmark' will run a new benchmarking selftest. 'bzr selftest --benchmark --lsprof-timed' will use lsprofile to generate profile data for the individual profiled calls, allowing for fine grained analysis of performance. (Robert Collins, Martin Pool).
- 'bzr commit' shows a progress bar. This is useful for commits over sftp where commit can take an appreciable time. (Robert Collins)
- 'bzr add' is now less verbose in telling you what ignore globs were matched by files being ignored. Instead it just tells you how many were ignored (because you might reasonably be expecting none to be ignored). 'bzr add -v' is unchanged and will report every ignored file. (Robert Collins).
- ftp now has a test server if medusa is installed. As part of testing, ftp support has been improved, including support for supplying a non-standard port. (John Arbash Meinel).
- 'bzr log --line' shows the revision number, and uses only the first line of the log message (#5162, Alexander Belchenko; Matthieu Moy)
- 'bzr status' has had the --all option removed. The 'bzr ls' command should be used to retrieve all versioned files. (Robert Collins)
- 'bzr bundle OTHER/BRANCH' will create a bundle which can be sent over email, and applied on the other end, while maintaining ancestry. This bundle can be applied with either 'bzr merge' or 'bzr pull', the same way you would apply another branch. (John Arbash Meinel, Aaron Bentley)
- 'bzr whoami' can now be used to set your identity from the command line, for a branch or globally. (Robey Pointer)
- 'bzr checkout' now aliased to 'bzr co', and 'bzr annotate' to 'bzr ann'. (Michael Ellerman)
- 'bzr revert DIRECTORY' now reverts the contents of the directory as well. (Aaron Bentley)
- 'bzr get sftp://foo' gives a better error when paramiko is not present. Also updates things like 'http+pycurl://' if pycurl is not present. (John Arbash Meinel) (Malone #47821, #52204)
- New env variable BZR_PROGRESS_BAR, sets the default progress bar type. Can be set to 'none' or 'dummy' to disable the progress bar, 'dots' or 'tty' to create the respective type. (John Arbash Meinel, #42197, #51107)
- Improve the help text for 'bzr diff' to explain what various options do. (John Arbash Meinel, #6391)
- 'bzr uncommit -r 10' now uncommits revisions 11.. rather than uncommitting revision 10. This makes -r10 more in line with what other commands do. 'bzr uncommit' also now saves the pending merges of the revisions that were removed. So it is safe to uncommit after a merge, fix something, and commit again. (John Arbash Meinel, #32526, #31426)
- 'bzr init' now also works on remote locations. (Wouter van Heyst, #48904)
- HTTP support has been updated. When using pycurl we now support connection keep-alive, which reduces dns requests and round trips. And for both urllib and pycurl we support multi-range requests, which decreases the number of round-trips. Performance results for bzr branch http://bazaar-vcs.org/bzr/bzr.dev/ indicate http branching is now 2-3x faster, and bzr pull in an existing branch is as much as 4x faster. (Michael Ellerman, Johan Rydberg, John Arbash Meinel, #46768)
- Performance improvements for sftp. Branching and pulling are now up to 2x faster. Utilize paramiko.readv() support for async requests if it is available (paramiko > 1.6) (John Arbash Meinel)
BUG FIXES:
- Fix shadowed definition of TestLocationConfig that caused some tests not to run. (Erik Bågfors, Michael Ellerman, Martin Pool, #32587)
- Fix unnecessary requirement of sign-my-commits that it be run from a working directory. (Martin Pool, Robert Collins)
- 'bzr push location' will only remember the push location if it succeeds in connecting to the remote location. (John Arbash Meinel, #49742)
- 'bzr revert' no longer toggles the executable bit on win32 (John Arbash Meinel, #45010)
- Handle broken pipe under win32 correctly. (John Arbash Meinel)
- sftp tests now work correctly on win32 if you have a newer paramiko (John Arbash Meinel)
- Cleanup win32 test suite, and general cleanup of places where file handles were being held open. (John Arbash Meinel)
- When specifying filenames for 'diff -r x..y', the name of the file in the working directory can be used, even if its name is different in both x and y.
- File-ids containing single- or double-quotes are handled correctly by push. (Aaron Bentley, #52227)
- Normalize unicode filenames to ensure cross-platform consistency. (John Arbash Meinel, #43689)
- The argument parser can now handle '-' as an argument. Currently no code interprets it specially (it is mostly handled as a file named '-'). But plugins, and future operations can use it. (John Arbash meinel, #50984)
- Bundles can properly read binary files with a plain 'r' in them. (John Arbash Meinel, #51927)
- Tuning iter_entries() to be more efficient (John Arbash Meinel, #5444)
- Lots of win32 fixes (the test suite passes again). (John Arbash Meinel, #50155)
- Handle openbsd returning None for sys.getfilesystemencoding() (#41183)
- Support ftp APPE (append) to allow Knits to be used over ftp (#42592)
- Removals are only committed if they match the filespec (or if there is no filespec). (#46635, Aaron Bentley)
- smart-add recurses through all supplied directories (John Arbash Meinel, #52578)
- Make the bundle reader extra lines before and after the bundle text. This allows you to parse an email with the bundle inline. (John Arbash Meinel, #49182)
- Change the file id generator to squash a little bit more. Helps when working with long filenames on windows. (Also helps for unicode filenames not generating hidden files). (John Arbash Meinel, #43801)
- Restore terminal mode on C-c while reading sftp password. (#48923, Nicholas Allen, Martin Pool)
- Timestamps are rounded to 1ms, and revision entries can be recreated exactly. (John Arbash Meinel, Jamie Wilkinson, #40693)
- Branch.base has changed to a URL, but ~/.bazaar/locations.conf should use local paths, since it is user visible (John Arbash Meinel, #53653)
- bzr status foo when foo was unversioned used to cause a full delta to be generated (John Arbash Meinel, #53638)
- When reading revision properties, an empty value should be considered the empty string, not None (John Arbash Meinel, #47782)
- bzr diff --diff-options can now handle binary files being changed. Also, the output is consistent when --diff-options is not supplied. (John Arbash Meinel, #54651, #52930)
- Use the right suffixes for loading plugins (John Arbash Meinel, #51810)
- Fix Branch.get_parent() to handle the case when the parent is not accessible (John Arbash Meinel, #52976)
INTERNALS:
- Combine the ignore rules into a single regex rather than looping over them to reduce the threshold where N^2 behaviour occurs in operations like status. (Jan Hudec, Robert Collins).
- Appending to bzrlib.DEFAULT_IGNORE is now deprecated. Instead, use one of the add functions in bzrlib.ignores. (John Arbash Meinel)
- 'bzr push' should only push the ancestry of the current revision, not all of the history in the repository. This is especially important for shared repositories. (John Arbash Meinel)
- bzrlib.delta.compare_trees now iterates in alphabetically sorted order, rather than randomly walking the inventories. (John Arbash Meinel)
- Doctests are now run in temporary directories which are cleaned up when they finish, rather than using special ScratchDir/ScratchBranch objects. (Martin Pool)
- Split check into separate methods on the branch and on the repository, so that it can be specialized in ways that are useful or efficient for different formats. (Martin Pool, Robert Collins)
- Deprecate Repository.all_revision_ids; most methods don't really need the global revision graph but only that part leading up to a particular revision. (Martin Pool, Robert Collins)
- Add a BzrDirFormat control_formats list which allows for control formats that do not use '.bzr' to store their data - i.e. '.svn', '.hg' etc. (Robert Collins, Jelmer Vernooij).
- bzrlib.diff.external_diff can be redirected to any file-like object. Uses subprocess instead of spawnvp. (James Henstridge, John Arbash Meinel, #4047, #48914)
- New command line option '--profile-imports', which will install a custom importer to log time to import modules and regex compilation time to sys.stderr (John Arbash Meinel)
- 'EmptyTree' is now deprecated, please use repository.revision_tree(None) instead. (Robert Collins)
- "RevisionTree" is now in bzrlib/revisiontree.py. (Robert Collins)
BUG FIXES:
- setup.py failed to install launchpad plugin. (Martin Pool)
BUG FIXES:
- Fix failure to commit a merge in a checkout. (Martin Pool, Robert Collins, Erik Bågfors, #43959)
- Nicer messages from 'commit' in the case of renames, and correct messages when a merge has occured. (Robert Collins, Martin Pool)
- Separate functionality from assert statements as they are skipped in optimized mode of python. Add the same check to pending merges. (Olaf Conradi, #44443)
CHANGES:
- Do not show the None revision in output of bzr ancestry. (Olaf Conradi)
- Add info on standalone branches without a working tree. (Olaf Conradi, #44155)
- Fix bug in knits when raising InvalidRevisionId. (Olaf Conradi, #44284)
CHANGES:
- Make editor invocation comply with Debian Policy. First check environment variables VISUAL and EDITOR, then try editor from alternatives system. If that all fails, fall back to the pre-defined list of editors. (Olaf Conradi, #42904)
NEW FEATURES:
- New 'register-branch' command registers a public branch into Launchpad.net, where it can be associated with bugs, etc. (Martin Pool, Bjorn Tillenius, Robert Collins)
INTERNALS:
- New public api in InventoryEntry - describe_change(old, new) which provides a human description of the changes between two old and new. (Robert Collins, Martin Pool)
TESTING:
- Fix test case for bzr info in upgrading a standalone branch to metadir, uses bzrlib api now. (Olaf Conradi)
NOTES WHEN UPGRADING:
Release 0.8 of bzr introduces a new format for history storage, called 'knit', as an evolution of to the 'weave' format used in 0.7. Local and remote operations are faster using knits than weaves. Several operations including 'init', 'init-repo', and 'upgrade' take a --format option that controls this. Branching from an existing branch will keep the same format.
It is possible to merge, pull and push between branches of different formats but this is slower than moving data between homogenous branches. It is therefore recommended (but not required) that you upgrade all branches for a project at the same time. Information on formats is shown by 'bzr info'.
bzr 0.8 now allows creation of 'repositories', which hold the history of files and revisions for several branches. Previously bzr kept all the history for a branch within the .bzr directory at the root of the branch, and this is still the default. To create a repository, use the new 'bzr init-repo' command. Branches exist as directories under the repository and contain just a small amount of information indicating the current revision of the branch.
bzr 0.8 also supports 'checkouts', which are similar to in cvs and subversion. Checkouts are associated with a branch (optionally in a repository), which contains all the historical information. The result is that a checkout can be deleted without losing any already-committed revisions. A new 'update' command is also available.
Repositories and checkouts are not supported with the 0.7 storage format. To use them you must upgrad to either knits, or to the 'metaweave' format, which uses weaves but changes the .bzr directory arrangement.
IMPROVEMENTS:
Sftp paths can now be relative, or local, according to the lftp convention. Paths now take the form:
sftp://user:pass@host:port/~/relative/path or sftp://user:pass@host:port/absolute/pathThe FTP transport now tries to reconnect after a temporary failure. ftp put is made atomic. (Matthieu Moy)
The FTP transport now maintains a pool of connections, and reuses them to avoid multiple connections to the same host (like sftp did). (Daniel Silverstone)
The bzr_man.py file has been removed. To create the man page now, use ./generate_docs.py man. The new program can also create other files. Run python generate_docs.py --help for usage information. (Hans Ulrich Niedermann & James Blackwell).
Man Page now gives full help (James Blackwell). Help also updated to reflect user config now being stored in .bazaar (Hans Ulrich Niedermann)
It's now possible to set aliases in bazaar.conf (Erik Bågfors)
Pull now accepts a --revision argument (Erik Bågfors)
bzr re-sign now allows multiple revisions to be supplied on the command line. You can now use the following command to sign all of your old commits:
find .bzr/revision-store// -name my@email-* \ | sed 's/.*\/\/..\///' \ | xargs bzr re-signUpgrade can now upgrade over the network. (Robert Collins)
Two new commands 'bzr checkout' and 'bzr update' allow for CVS/SVN-alike behaviour. By default they will cache history in the checkout, but with --lightweight almost all data is kept in the master branch. (Robert Collins)
'revert' unversions newly-versioned files, instead of deleting them.
'merge' is more robust. Conflict messages have changed.
'merge' and 'revert' no longer clobber existing files that end in '~' or '.moved'.
Default log format can be set in configuration and plugins can register their own formatters. (Erik Bågfors)
New 'reconcile' command will check branch consistency and repair indexes that can become out of sync in pre 0.8 formats. (Robert Collins, Daniel Silverstone)
New 'bzr init --format' and 'bzr upgrade --format' option to control what storage format is created or produced. (Robert Collins, Martin Pool)
Add parent location to 'bzr info', if there is one. (Olaf Conradi)
New developer commands 'weave-list' and 'weave-join'. (Martin Pool)
New 'init-repository' command, plus support for repositories in 'init' and 'branch' (Aaron Bentley, Erik Bågfors, Robert Collins)
Improve output of 'info' command. Show all relevant locations related to working tree, branch and repository. Use kibibytes for binary quantities. Fix off-by-one error in missing revisions of working tree. Make 'info' work on branches, repositories and remote locations. Show locations relative to the shared repository, if applicable. Show locking status of locations. (Olaf Conradi)
Diff and merge now safely handle binary files. (Aaron Bentley)
'pull' and 'push' now normalise the revision history, so that any two branches with the same tip revision will have the same output from 'log'. (Robert Collins)
'merge' accepts --remember option to store parent location, like 'push' and 'pull'. (Olaf Conradi)
bzr status and diff when files given as arguments do not exist in the relevant trees. (Martin Pool, #3619)
Add '.hg' to the default ignore list. (Martin Pool)
'knit' is now the default disk format. This improves disk performance and utilization, increases incremental pull performance, robustness with SFTP and allows checkouts over SFTP to perform acceptably. The initial Knit code was contributed by Johan Rydberg based on a specification by Martin Pool. (Robert Collins, Aaron Bentley, Johan Rydberg, Martin Pool).
New tool to generate all-in-one html version of the manual. (Alexander Belchenko)
Hitting CTRL-C while doing an SFTP push will no longer cause stale locks to be left in the SFTP repository. (Robert Collins, Martin Pool).
New option 'diff --prefix' to control how files are named in diff output, with shortcuts '-p0' and '-p1' corresponding to the options for GNU patch. (Alexander Belchenko, Goffredo Baroncelli, Martin Pool)
Add --revision option to 'annotate' command. (Olaf Conradi)
If bzr shows an unexpected revision-history after pulling (perhaps due to a reweave) it can now be corrected by 'bzr reconcile'. (Robert Collins)
CHANGES:
- Commit is now verbose by default, and shows changed filenames and the new revision number. (Robert Collins, Martin Pool)
- Unify 'mv', 'move', 'rename'. (Matthew Fuller, #5379)
- 'bzr -h' shows help. (Martin Pool, Ian Bicking, #35940)
- Make 'pull' and 'push' remember location on failure using --remember. (Olaf Conradi)
- For compatibility, make old format for using weaves inside metadir available as 'metaweave' format. Rename format 'metadir' to 'default'. Clean up help for option --format in commands 'init', 'init-repo' and 'upgrade'. (Olaf Conradi)
INTERNALS:
- The internal storage of history, and logical branch identity have now been split into Branch, and Repository. The common locking and file management routines are now in bzrlib.lockablefiles. (Aaron Bentley, Robert Collins, Martin Pool)
- Transports can now raise DependencyNotPresent if they need a library which is not installed, and then another implementation will be tried. (Martin Pool)
- Remove obsolete (and no-op) decode parameter to Transport.get. (Martin Pool)
- Using Tree Transform for merge, revert, tree-building
- WorkingTree.create, Branch.create, WorkingTree.create_standalone, Branch.initialize are now deprecated. Please see BzrDir.create_* for replacement API's. (Robert Collins)
- New BzrDir class represents the .bzr control directory and manages formatting issues. (Robert Collins)
- New repository.InterRepository class encapsulates Repository to Repository actions and allows for clean selection of optimised code paths. (Robert Collins)
- bzrlib.fetch.fetch and bzrlib.fetch.greedy_fetch are now deprecated, please use branch.fetch or repository.fetch depending on your needs. (Robert Collins)
- deprecated methods now have a is_deprecated flag on them that can be checked, if you need to determine whether a given callable is deprecated at runtime. (Robert Collins)
- Progress bars are now nested - see bzrlib.ui.ui_factory.nested_progress_bar. (Robert Collins, Robey Pointer)
- New API call get_format_description() for each type of format. (Olaf Conradi)
- Changed branch.set_parent() to accept None to remove parent. (Olaf Conradi)
- Deprecated BzrError AmbiguousBase. (Olaf Conradi)
- WorkingTree.branch is now a read only property. (Robert Collins)
- bzrlib.ui.text.TextUIFactory now accepts a bar_type parameter which can be None or a factory that will create a progress bar. This is useful for testing or for overriding the bzrlib.progress heuristic. (Robert Collins)
- New API method get_physical_lock_status() to query locks present on a transport. (Olaf Conradi)
- Repository.reconcile now takes a thorough keyword parameter to allow requesting an indepth reconciliation, rather than just a data-loss check. (Robert Collins)
- bzrlib.ui.ui_factory protocol now supports get_boolean to prompt the user for yes/no style input. (Robert Collins)
TESTING:
- SFTP tests now shortcut the SSH negotiation, reducing test overhead for testing SFTP protocol support. (Robey Pointer)
- Branch formats are now tested once per implementation (see bzrlib. tests.branch_implementations. This is analagous to the transport interface tests, and has been followed up with working tree, repository and BzrDir tests. (Robert Collins)
- New test base class TestCaseWithTransport provides a transport aware test environment, useful for testing any transport-interface using code. The test suite option --transport controls the transport used by this class (when its not being used as part of implementation contract testing). (Robert Collins)
- Close logging handler on disabling the test log. This will remove the handler from the internal list inside python's logging module, preventing shutdown from closing it twice. (Olaf Conradi)
- Move test case for uncommit to blackbox tests. (Olaf Conradi)
- run_bzr and run_bzr_captured now accept a 'stdin="foo"' parameter which will provide String("foo") to the command as its stdin.
CHANGES:
- .bzrignore is excluded from exports, on the grounds that it's a bzr internal-use file and may not be wanted. (Jamie Wilkinson)
- The "bzr directories" command were removed in favor of the new --kind option to the "bzr inventory" command. To list all versioned directories, now use "bzr inventory --kind directory". (Johan Rydberg)
- Under Windows configuration directory is now %APPDATA%\bazaar\2.0 by default. (John Arbash Meinel)
- The parent of Bzr configuration directory can be set by BZR_HOME environment variable. Now the path for it is searched in BZR_HOME, then in HOME. Under Windows the order is: BZR_HOME, APPDATA (usually points to C:\Documents and Settings\User Name\Application Data), HOME. (John Arbash Meinel)
- Plugins with the same name in different directories in the bzr plugin path are no longer loaded: only the first successfully loaded one is used. (Robert Collins)
- Use systems' external ssh command to open connections if possible. This gives better integration with user settings such as ProxyCommand. (James Henstridge)
- Permissions on files underneath .bzr/ are inherited from the .bzr directory. So for a shared repository, simply doing 'chmod -R g+w .bzr/' will mean that future file will be created with group write permissions.
- configure.in and config.guess are no longer in the builtin default ignore list.
- '.sw[nop]' pattern ignored, to ignore vim swap files for nameless files. (John Arbash Meinel, Martin Pool)
IMPROVEMENTS:
- "bzr INIT dir" now initializes the specified directory, and creates it if it does not exist. (John Arbash Meinel)
- New remerge command (Aaron Bentley)
- Better zsh completion script. (Steve Borho)
- 'bzr diff' now returns 1 when there are changes in the working tree. (Robert Collins)
- 'bzr push' now exists and can push changes to a remote location. This uses the transport infrastructure, and can store the remote location in the ~/.bazaar/branches.conf configuration file. (Robert Collins)
- Test directories are only kept if the test fails and the user requests that they be kept.
- Tweaks to short log printing
- Added branch nicks, new nick command, printing them in log output. (Aaron Bentley)
- If $BZR_PDB is set, pop into the debugger when an uncaught exception occurs. (Martin Pool)
- Accept 'bzr resolved' (an alias for 'bzr resolve'), as this is the same as Subversion. (Martin Pool)
- New ftp transport support (on ftplib), for ftp:// and aftp:// URLs. (Daniel Silverstone)
- Commit editor temporary files now start with bzr_log., to allow text editors to match the file name and set up appropriate modes or settings. (Magnus Therning)
- Improved performance when integrating changes from a remote weave. (Goffredo Baroncelli)
- Sftp will attempt to cache the connection, so it is more likely that a connection will be reused, rather than requiring multiple password requests.
- bzr revno now takes an optional argument indicating the branch whose revno should be printed. (Michael Ellerman)
- bzr cat defaults to printing the last version of the file. (Matthieu Moy, #3632)
- New global option 'bzr --lsprof COMMAND' runs bzr under the lsprof profiler. (Denys Duchier)
- Faster commits by reading only the headers of affected weave files. (Denys Duchier)
- 'bzr add' now takes a --dry-run parameter which shows you what would be added, but doesn't actually add anything. (Michael Ellerman)
- 'bzr add' now lists how many files were ignored per glob. add --verbose lists the specific files. (Aaron Bentley)
- 'bzr missing' now supports displaying changes in diverged trees and can be limited to show what either end of the comparison is missing. (Aaron Bently, with a little prompting from Daniel Silverstone)
BUG FIXES:
- SFTP can walk up to the root path without index errors. (Robert Collins)
- Fix bugs in running bzr with 'python -O'. (Martin Pool)
- Error when run with -OO
- Fix bug in reporting http errors that don't have an http error code. (Martin Pool)
- Handle more cases of pipe errors in display commands
- Change status to 3 for all errors
- Files that are added and unlinked before committing are completely ignored by diff and status
- Stores with some compressed texts and some uncompressed texts are now able to be used. (John A Meinel)
- Fix for bzr pull failing sometimes under windows
- Fix for sftp transport under windows when using interactive auth
- Show files which are both renamed and modified as such in 'bzr status' output. (Daniel Silverstone, #4503)
- Make annotate cope better with revisions committed without a valid email address. (Marien Zwart)
- Fix representation of tab characters in commit messages. (Harald Meland)
- List of plugin directories in BZR_PLUGIN_PATH environment variable is now parsed properly under Windows. (Alexander Belchenko)
- Show number of revisions pushed/pulled/merged. (Robey Pointer)
- Keep a cached copy of the basis inventory to speed up operations that need to refer to it. (Johan Rydberg, Martin Pool)
- Fix bugs in bzr status display of non-ascii characters. (Martin Pool)
- Remove Makefile.in from default ignore list. (Tollef Fog Heen, Martin Pool, #6413)
- Fix failure in 'bzr added'. (Nathan McCallum, Martin Pool)
TESTING:
- Fix selftest asking for passwords when there are no SFTP keys. (Robey Pointer, Jelmer Vernooij)
- Fix selftest run with 'python -O'. (Martin Pool)
- Fix HTTP tests under Windows. (John Arbash Meinel)
- Make tests work even if HOME is not set (Aaron Bentley)
- Updated build_tree to use fixed line-endings for tests which read the file cotents and compare. Make some tests use this to pass under Windows. (John Arbash Meinel)
- Skip stat and symlink tests under Windows. (Alexander Belchenko)
- Delay in selftest/testhashcash is now issued under win32 and Cygwin. (John Arbash Meinel)
- Use terminal width to align verbose test output. (Martin Pool)
- Blackbox tests are maintained within the bzrlib.tests.blackbox directory. If adding a new test script please add that to bzrlib.tests.blackbox.__init__. (Robert Collins)
- Much better error message if one of the test suites can't be imported. (Martin Pool)
- Make check now runs the test suite twice - once with the default locale, and once with all locales forced to C, to expose bugs. This is not trivially done within python, so for now its only triggered by running Make check. Integrators and packagers who wish to check for full platform support should run 'make check' to test the source. (Robert Collins)
- Tests can now run TestSkipped if they can't execute for any reason. (Martin Pool) (NB: TestSkipped should only be raised for correctable reasons - see the wiki spec ImprovingBzrTestSuite).
- Test sftp with relative, absolute-in-homedir and absolute-not-in-homedir paths for the transport tests. Introduce blackbox remote sftp tests that test the same permutations. (Robert Collins, Robey Pointer)
- Transport implementation tests are now independent of the local file system, which allows tests for esoteric transports, and for features not available in the local file system. They also repeat for variations on the URL scheme that can introduce issues in the transport code, see bzrlib.transport.TransportTestProviderAdapter() for this. (Robert Collins).
- TestCase.build_tree uses the transport interface to build trees, pass in a transport parameter to give it an existing connection. (Robert Collins).
INTERNALS:
WorkingTree.pull has been split across Branch and WorkingTree, to allow Branch only pulls. (Robert Collins)
commands.display_command now returns the result of the decorated function. (Robert Collins)
LocationConfig now has a set_user_option(key, value) call to save a setting in its matching location section (a new one is created if needed). (Robert Collins)
Branch has two new methods, get_push_location and set_push_location to respectively, get and set the push location. (Robert Collins)
commands.register_command now takes an optional flag to signal that the registrant is planning to decorate an existing command. When given multiple plugins registering a command is not an error, and the original command class (whether built in or a plugin based one) is returned to the caller. There is a new error 'MustUseDecorated' for signalling when a wrapping command should switch to the original version. (Robert Collins)
Some option parsing errors will raise 'BzrOptionError', allowing granular detection for decorating commands. (Robert Collins).
Branch.read_working_inventory has moved to WorkingTree.read_working_inventory. This necessitated changes to Branch.get_root_id, and a move of Branch.set_inventory to WorkingTree as well. To make it clear that a WorkingTree cannot always be obtained Branch.working_tree() will raise errors.NoWorkingTree if one cannot be obtained. (Robert Collins)
All pending merges operations from Branch are now on WorkingTree. (Robert Collins)
The follow operations from Branch have moved to WorkingTree:
add() commit() move() rename_one() unknowns()(Robert Collins)
bzrlib.add.smart_add_branch is now smart_add_tree. (Robert Collins)
New "rio" serialization format, similar to rfc-822. (Martin Pool)
Rename selftests to bzrlib.tests.test_foo. (John A Meinel, Martin Pool)
bzrlib.plugin.all_plugins has been changed from an attribute to a query method. (Robert Collins)
New options to read only the table-of-contents of a weave. (Denys Duchier)
Raise NoSuchFile when someone tries to add a non-existant file. (Michael Ellerman)
Simplify handling of DivergedBranches in cmd_pull(). (Michael Ellerman)
Branch.controlfile* logic has moved to lockablefiles.LockableFiles, which is exposed as Branch().control_files. Also this has been altered with the controlfile pre/suffix replaced by simple method names like 'get' and 'put'. (Aaron Bentley, Robert Collins).
Deprecated functions and methods can now be marked as such using the bzrlib.symbol_versioning module. Marked method have their docstring updated and will issue a DeprecationWarning using the warnings module when they are used. (Robert Collins)
bzrlib.osutils.safe_unicode now exists to provide parameter coercion for functions that need unicode strings. (Robert Collins)
IMPROVEMENTS:
- pull now takes --verbose to show you what revisions are added or removed (John A Meinel)
- merge now takes a --show-base option to include the base text in conflicts. (Aaron Bentley)
- The config files are now read using ConfigObj, so '=' should be used as a separator, not ':'. (Aaron Bentley)
- New 'bzr commit --strict' option refuses to commit if there are any unknown files in the tree. To commit, make sure all files are either ignored, added, or deleted. (Michael Ellerman)
- The config directory is now ~/.bazaar, and there is a single file ~/.bazaar/bazaar.conf storing email, editor and other preferences. (Robert Collins)
- 'bzr add' no longer takes a --verbose option, and a --quiet option has been added that suppresses all output.
- Improved zsh completion support in contrib/zsh, from Clint Adams.
- Builtin 'bzr annotate' command, by Martin Pool with improvements from Goffredo Baroncelli.
- 'bzr check' now accepts -v for verbose reporting, and checks for ghosts in the branch. (Robert Collins)
- New command 're-sign' which will regenerate the gpg signature for a revision. (Robert Collins)
- If you set check_signatures=require for a path in ~/.bazaar/branches.conf then bzr will invoke your gpg_signing_command (defaults to gpg) and record a digital signature of your commit. (Robert Collins)
- New sftp transport, based on Paramiko. (Robey Pointer)
- 'bzr pull' now accepts '--clobber' which will discard local changes and make this branch identical to the source branch. (Robert Collins)
- Just give a quieter warning if a plugin can't be loaded, and put the details in .bzr.log. (Martin Pool)
- 'bzr branch' will now set the branch-name to the last component of the output directory, if one was supplied.
- If the option post_commit is set to one (or more) python function names (must be in the bzrlib namespace), then they will be invoked after the commit has completed, with the branch and revision_id as parameters. (Robert Collins)
- Merge now has a retcode of 1 when conflicts occur. (Robert Collins)
- --merge-type weave is now supported for file contents. Tree-shape changes are still three-way based. (Martin Pool, Aaron Bentley)
- 'bzr check' allows the first revision on revision-history to have parents - something that is expected for cheap checkouts, and occurs when conversions from baz do not have all history. (Robert Collins).
- 'bzr merge' can now graft unrelated trees together, if your specify 0 as a base. (Aaron Bentley)
- 'bzr commit branch' and 'bzr commit branch/file1 branch/file2' now work (Aaron Bentley)
- Add '.sconsign*' to default ignore list. (Alexander Belchenko)
- 'bzr merge --reprocess' minimizes conflicts
TESTING:
- The 'bzr selftest --pattern' option for has been removed, now test specifiers on the command line can be simple strings, or regexps, or both. (Robert Collins)
- Passing -v to selftest will now show the time each test took to complete, which will aid in analysing performance regressions and related questions. (Robert Collins)
- 'bzr selftest' runs all tests, even if one fails, unless '--one' is given. (Martin Pool)
- There is a new method for TestCaseInTempDir, assertFileEqual, which will check that a given content is equal to the content of the named file. (Robert Collins)
- Fix test suite's habit of leaving many temporary log files in $TMPDIR. (Martin Pool)
INTERNALS:
- New 'testament' command and concept for making gpg-signatures of revisions that are not tied to a particular internal representation. (Martin Pool).
- Per-revision properties ('revprops') as key-value associated strings on each revision created when the revision is committed. Intended mainly for the use of external tools. (Martin Pool).
- Config options have moved from bzrlib.osutils to bzrlib.config. (Robert Collins)
- Improved command line option definitions allowing explanations for individual options, among other things. Contributed by Magnus Therning.
- Config options have moved from bzrlib.osutils to bzrlib.config. Configuration is now done via the config.Config interface: Depending on whether you have a Branch, a Location or no information available, construct a *Config, and use its signature_checking, username and user_email methods. (Robert Collins)
- Plugins are now loaded under bzrlib.plugins, not bzrlib.plugin, and they are made available for other plugins to use. You should not import other plugins during the __init__ of your plugin though, as no ordering is guaranteed, and the plugins directory is not on the python path. (Robert Collins)
- Branch.relpath has been moved to WorkingTree.relpath. WorkingTree no no longer takes an inventory, rather it takes an option branch parameter, and if None is given will open the branch at basedir implicitly. (Robert Collins)
- Cleaner exception structure and error reporting. Suggested by Scott James Remnant. (Martin Pool)
- Branch.remove has been moved to WorkingTree, which has also gained lock_read, lock_write and unlock methods for convenience. (Robert Collins)
- Two decorators, needs_read_lock and needs_write_lock have been added to the branch module. Use these to cause a function to run in a read or write lock respectively. (Robert Collins)
- Branch.open_containing now returns a tuple (Branch, relative-path), which allows direct access to the common case of 'get me this file from its branch'. (Robert Collins)
- Transports can register using register_lazy_transport, and they will be loaded when first used. (Martin Pool)
- 'pull' has been factored out of the command as WorkingTree.pull(). A new option to WorkingTree.pull has been added, clobber, which will ignore diverged history and pull anyway. (Robert Collins)
- config.Config has a get_user_option call that accepts an option name. This will be looked up in branches.conf and bazaar.conf as normal. It is intended that this be used by plugins to support options - options of built in programs should have specific methods on the config. (Robert Collins)
- merge.merge_inner now has tempdir as an optional parameter. (Robert Collins)
- Tree.kind is not recorded at the top level of the hierarchy, as it was missing on EmptyTree, leading to a bug with merge on EmptyTrees. (Robert Collins)
- WorkingTree.__del__ has been removed, it was non deterministic and not doing what it was intended to. See WorkingTree.__init__ for a comment about future directions. (Robert Collins/Martin Pool)
- bzrlib.transport.http has been modified so that only 404 urllib errors are returned as NoSuchFile. Other exceptions will propogate as normal. This allows debuging of actual errors. (Robert Collins)
- bzrlib.transport.Transport now accepts ONLY url escaped relative paths to apis like 'put', 'get' and 'has'. This is to provide consistent behaviour - it operates on url's only. (Robert Collins)
- Transports can register using register_lazy_transport, and they will be loaded when first used. (Martin Pool)
- merge_flex no longer calls conflict_handler.finalize(), instead that is called by merge_inner. This is so that the conflict count can be retrieved (and potentially manipulated) before returning to the caller of merge_inner. Likewise 'merge' now returns the conflict count to the caller. (Robert Collins)
- revision.revision_graph can handle having only partial history for a revision - that is no revisions in the graph with no parents. (Robert Collins).
- New builtins.branch_files uses the standard file_list rules to produce a branch and a list of paths, relative to that branch (Aaron Bentley)
- New TestCase.addCleanup facility.
- New bzrlib.version_info tuple (similar to sys.version_info), which can be used by programs importing bzrlib.
BUG FIXES:
- Better handling of branches in directories with non-ascii names. (Joel Rosdahl, Panagiotis Papadakos)
- Upgrades of trees with no commits will not fail due to accessing [-1] in the revision-history. (Andres Salomon)
BUG FIXES:
- Fix problem in pulling over http from machines that do not allow directories to be listed.
- Avoid harmless warning about invalid hash cache after upgrading branch format.
PERFORMANCE:
- Avoid some unnecessary http operations in branch and pull.
NOTES:
- 'bzr branch' over http initially gives a very high estimate of completion time but it should fall as the first few revisions are pulled in. branch is still slow on high-latency connections.
BUG FIXES:
- bzr-man.py has been updated to work again. Contributed by Rob Weir.
- Locking is now done with fcntl.lockf which works with NFS file systems. Contributed by Harald Meland.
- When a merge encounters a file that has been deleted on one side and modified on the other, the old contents are written out to foo.BASE and foo.SIDE, where SIDE is this or OTHER. Contributed by Aaron Bentley.
- Export was choosing incorrect file paths for the content of the tarball, this has been fixed by Aaron Bentley.
- Commit will no longer commit without a log message, an error is returned instead. Contributed by Jelmer Vernooij.
- If you commit a specific file in a sub directory, any of its parent directories that are added but not listed will be automatically included. Suggested by Michael Ellerman.
- bzr commit and upgrade did not correctly record new revisions for files with only a change to their executable status. bzr will correct this when it encounters it. Fixed by Robert Collins
- HTTP tests now force off the use of http_proxy for the duration. Contributed by Gustavo Niemeyer.
- Fix problems in merging weave-based branches that have different partial views of history.
- Symlink support: working with symlinks when not in the root of a bzr tree was broken, patch from Scott James Remnant.
IMPROVEMENTS:
- 'branch' now accepts a --basis parameter which will take advantage of local history when making a new branch. This allows faster branching of remote branches. Contributed by Aaron Bentley.
- New tree format based on weave files, called version 5. Existing branches can be upgraded to this format using 'bzr upgrade'.
- Symlinks are now versionable. Initial patch by Erik Toubro Nielsen, updated to head by Robert Collins.
- Executable bits are tracked on files. Patch from Gustavo Niemeyer.
- 'bzr status' now shows unknown files inside a selected directory. Patch from Heikki Paajanen.
- Merge conflicts are recorded in .bzr. Two new commands 'conflicts' and 'resolve' have needed added, which list and remove those merge conflicts respectively. A conflicted tree cannot be committed in. Contributed by Aaron Bentley.
- 'rm' is now an alias for 'remove'.
- Stores now split out their content in a single byte prefixed hash, dropping the density of files per directory by 256. Contributed by Gustavo Niemeyer.
- 'bzr diff -r branch:URL' will now perform a diff between two branches. Contributed by Robert Collins.
- 'bzr log' with the default formatter will show merged revisions, indented to the right. Initial implementation contributed by Gustavo Niemeyer, made incremental by Robert Collins.
INTERNALS:
- Test case failures have the exception printed after the log for your viewing pleasure.
- InventoryEntry is now an abstract base class, use one of the concrete InventoryDirectory etc classes instead.
- Branch raises an UnsupportedFormatError when it detects a bzr branch it cannot understand. This allows for precise handling of such circumstances.
- Remove RevisionReference class; Revision.parent_ids is now simply a list of their ids and parent_sha1s is a list of their corresponding sha1s (for old branches only at the moment.)
- New method-object style interface for Commit() and Fetch().
- Renamed Branch.last_patch() to Branch.last_revision(), since we call them revisions not patches.
- Move copy_branch to bzrlib.clone.copy_branch. The destination directory is created if it doesn't exist.
- Inventories now identify the files which were present by giving the revision of that file.
- Inventory and Revision XML contains a version identifier. This must be consistent with the overall branch version but allows for more flexibility in future upgrades.
TESTING:
- Removed testsweet module so that tests can be run after bzr installed by 'bzr selftest'.
- 'bzr selftest' command-line arguments can now be partial ids of tests to run, e.g. bzr selftest test_weave
BUG FIXES:
- Fixed "branch -r" option.
- Fix remote access to branches containing non-compressed history. (Robert Collins).
- Better reliability of http server tests. (John Arbash-Meinel)
- Merge graph maximum distance calculation fix. (Aaron Bentley)
- Various minor bug in windows support have been fixed, largely in the test suite. Contributed by Alexander Belchenko.
IMPROVEMENTS:
- Status now accepts a -r argument to give status between chosen revisions. Contributed by Heikki Paajanen.
- Revision arguments no longer use +/-/= to control ranges, instead there is a 'before' namespace, which limits the successive namespace. For example '$ bzr log -r date:yesterday..before:date:today' will select everything from yesterday and before today. Contributed by Robey Pointer
- There is now a bzr.bat file created by distutils when building on Windows. Contributed by Alexander Belchenko.
INTERNALS:
- Removed uuid() as it was unused.
- Improved 'fetch' code for pulling revisions from one branch into another (used by pull, merged, etc.)
IMPROVEMENTS:
- Adding a file whose parent directory is not versioned will implicitly add the parent, and so on up to the root. This means you should never need to explictly add a directory, they'll just get added when you add a file in the directory. Contributed by Michael Ellerman.
- Ignore .DS_Store (contains Mac metadata) by default. (Nir Soffer)
- If you set BZR_EDITOR in the environment, it is checked in preference to EDITOR and the config file for the interactive commit editing program. Related to this is a bugfix where a missing program set in EDITOR would cause editing to fail, now the fallback program for the operating system is still tried.
- Files that are not directories/symlinks/regular files will no longer cause bzr to fail, it will just ignore them by default. You cannot add them to the tree though - they are not versionable.
INTERNALS:
- Refactor xml packing/unpacking.
BUG FIXES:
- Fixed 'bzr mv' by Ollie Rutherfurd.
- Fixed strange error when trying to access a nonexistent http branch.
- Make sure that the hashcache gets written out if it can't be read.
PORTABILITY:
- Various Windows fixes from Ollie Rutherfurd.
- Quieten warnings about locking; patch from Matt Lavin.
NEW FEATURES:
- bzr shell-complete command contributed by Clint Adams to help with intelligent shell completion.
- New expert command bzr find-merge-base for debugging merges.
ENHANCEMENTS:
- Much better merge support.
- merge3 conflicts are now reported with markers like '<<<<<<<' (seven characters) which is the same as CVS and pleases things like emacs smerge.
BUG FIXES:
- bzr upgrade no longer fails when trying to fix trees that mention revisions that are not present.
- Fixed bugs in listing plugins from bzr plugins.
- Fix case of $EDITOR containing options for the editor.
- Fix log -r refusing to show the last revision. (Patch from Goffredo Baroncelli.)
CHANGES:
- bzr log --show-ids shows the revision ids of all parents.
- Externally provided commands on your $BZRPATH no longer need to recognize --bzr-usage to work properly, and can just handle --help themselves.
LIBRARY:
- Changed trace messages to go through the standard logging framework, so that they can more easily be redirected by libraries.
NEW FEATURES:
- Python plugins, automatically loaded from the directories on BZR_PLUGIN_PATH or ~/.bzr.conf/plugins by default.
- New 'bzr mkdir' command.
- Commit mesage is fetched from an editor if not given on the command line; patch from Torsten Marek.
- bzr log -m FOO displays commits whose message matches regexp FOO.
- bzr add with no arguments adds everything under the current directory.
- bzr mv does move or rename depending on its arguments, like the Unix command.
- bzr missing command shows a summary of the differences between two trees. (Merged from John Arbash-Meinel.)
- An email address for commits to a particular tree can be specified by putting it into .bzr/email within a branch. (Based on a patch from Heikki Paajanen.)
ENHANCEMENTS:
- Faster working tree operations.
CHANGES:
- 3rd-party modules shipped with bzr are copied within the bzrlib python package, so that they can be installed by the setup script without clashing with anything already existing on the system. (Contributed by Gustavo Niemeyer.)
- Moved plugins directory to bzrlib/, so that there's a standard plugin directory which is not only installed with bzr itself but is also available when using bzr from the development tree. BZR_PLUGIN_PATH and DEFAULT_PLUGIN_PATH are then added to the standard plugins directory.
- When exporting to a tarball with bzr export --format tgz, put everything under a top directory rather than dumping it into the current directory. This can be overridden with the --root option. Patch from William Dodé and John Meinel.
- New bzr upgrade command to upgrade the format of a branch, replacing bzr check --update.
- Files within store directories are no longer marked readonly on disk.
- Changed bzr log output to a more compact form suggested by John A Meinel. Old format is available with the --long or -l option, patched by William Dodé.
- By default the commit command refuses to record a revision with no changes unless the --unchanged option is given.
- The --no-plugins, --profile and --builtin command line options must come before the command name because they affect what commands are available; all other options must come after the command name because their interpretation depends on it.
- branch and clone added as aliases for branch.
- Default log format is back to the long format; the compact one is available with --short.
BUG FIXES:
- Fix bugs in committing only selected files or within a subdirectory.
CHANGES:
- bzr with no command now shows help rather than giving an error. Suggested by Michael Ellerman.
- bzr status output format changed, because svn-style output doesn't really match the model of bzr. Now files are grouped by status and can be shown with their IDs. bzr status --all shows all versioned files and unknown files but not ignored files.
- bzr log runs from most-recent to least-recent, the reverse of the previous order. The previous behaviour can be obtained with the --forward option.
- bzr inventory by default shows only filenames, and also ids if --show-ids is given, in which case the id is the second field.
ENHANCEMENTS:
- New 'bzr whoami --email' option shows only the email component of the user identification, from Jo Vermeulen.
- New bzr ignore PATTERN command.
- Nicer error message for broken pipe, interrupt and similar conditions that don't indicate an internal error.
- Add .*.sw[nop] .git .*.tmp *,v to default ignore patterns.
- Per-branch locks keyed on .bzr/branch-lock, available in either read or write mode.
- New option bzr log --show-ids shows revision and file ids.
- New usage bzr log FILENAME shows only revisions that affected that file.
- Changed format for describing changes in bzr log -v.
- New option bzr commit --file to take a message from a file, suggested by LarstiQ.
- New syntax bzr status [FILE...] contributed by Bartosz Oler. File may be in a branch other than the working directory.
- bzr log and bzr root can be given an http URL instead of a filename.
- Commands can now be defined by external programs or scripts in a directory on $BZRPATH.
- New "stat cache" avoids reading the contents of files if they haven't changed since the previous time.
- If the Python interpreter is too old, try to find a better one or give an error. Based on a patch from Fredrik Lundh.
- New optional parameter bzr info [BRANCH].
- New form bzr commit SELECTED to commit only selected files.
- New form bzr log -r FROM:TO shows changes in selected range; contributed by John A Meinel.
- New option bzr diff --diff-options 'OPTS' allows passing options through to an external GNU diff.
- New option bzr add --no-recurse to add a directory but not their contents.
- bzr --version now shows more information if bzr is being run from a branch.
BUG FIXES:
- Fixed diff format so that added and removed files will be handled properly by patch. Fix from Lalo Martins.
- Various fixes for files whose names contain spaces or other metacharacters.
TESTING:
- Converted black-box test suites from Bourne shell into Python; now run using ./testbzr. Various structural improvements to the tests.
- testbzr by default runs the version of bzr found in the same directory as the tests, or the one given as the first parameter.
- testbzr also runs the internal tests, so the only command required to check is just ./testbzr.
- testbzr requires python2.4, but can be used to test bzr running under a different version.
- Tests added for many other changes in this release.
INTERNAL:
- Included ElementTree library upgraded to 1.2.6 by Fredrik Lundh.
- Refactor command functions into Command objects based on HCT by Scott James Remnant.
- Better help messages for many commands.
- Expose bzrlib.open_tracefile() to start the tracefile; until this is called trace messages are just discarded.
- New internal function find_touching_revisions() and hidden command touching-revisions trace the changes to a given file.
- Simpler and faster compare_inventories() function.
- bzrlib.open_tracefile() takes a tracefilename parameter.
- New AtomicFile class.
- New developer commands added, modified.
PORTABILITY:
- Cope on Windows on python2.3 by using the weaker random seed. 2.4 is now only recommended.
ENHANCEMENTS:
- 'bzr diff' optionally takes a list of files to diff. Still a bit basic. Patch from QuantumG.
- More default ignore patterns.
- New 'bzr log --verbose' shows a list of files changed in the changeset. Patch from Sebastian Cote.
- Roll over ~/.bzr.log if it gets too large.
- Command abbreviations 'ci', 'st', 'stat', '?' based on a patch by Jason Diamon.
- New 'bzr help commands' based on a patch from Denys Duchier.
CHANGES:
- User email is determined by looking at $BZREMAIL or ~/.bzr.email or $EMAIL. All are decoded by the locale preferred encoding. If none of these are present user@hostname is used. The host's fully-qualified name is not used because that tends to fail when there are DNS problems.
- New 'bzr whoami' command instead of username user-email.
BUG FIXES:
- Make commit safe for hardlinked bzr trees.
- Some Unicode/locale fixes.
- Partial workaround for difflib.unified_diff not handling trailing newlines properly.
INTERNAL:
- Allow docstrings for help to be in PEP0257 format. Patch from Matt Brubeck.
- More tests in test.sh.
- Write profile data to a temporary file not into working directory and delete it when done.
- Smaller .bzr.log with process ids.
PORTABILITY:
- Fix opening of ~/.bzr.log on Windows. Patch from Andrew Bennetts.
- Some improvements in handling paths on Windows, based on a patch from QuantumG.
ENHANCEMENTS:
- New "directories" internal command lists versioned directories in the tree.
- Can now say "bzr commit --help".
- New "rename" command to rename one file to a different name and/or directory.
- New "move" command to move one or more files into a different directory.
- New "renames" command lists files renamed since base revision.
- New cat command contributed by janmar.
CHANGES:
- .bzr.log is placed in $HOME (not pwd) and is always written in UTF-8. (Probably not a completely good long-term solution, but will do for now.)
PORTABILITY:
- Workaround for difflib bug in Python 2.3 that causes an exception when comparing empty files. Reported by Erik Toubro Nielsen.
INTERNAL:
- Refactored inventory storage to insert a root entry at the top.
TESTING:
- Start of shell-based black-box testing in test.sh.
PORTABILITY:
- Win32 fixes from Steve Brown.
ENHANCEMENTS:
- Default ignore list extended (see bzrlib/__init__.py).
- Patterns in .bzrignore are now added to the default ignore list, rather than replacing it.
- Ignore list isn't reread for every file.
- More help topics.
- Reinstate the 'bzr check' command to check invariants of the branch.
- New 'ignored' command lists which files are ignored and why; 'deleted' lists files deleted in the current working tree.
- Performance improvements.
- New global --profile option.
- Ignore patterns like './config.h' now correctly match files in the root directory only.
ENHANCEMENTS:
- More information from info command.
- Can now say "bzr help COMMAND" for more detailed help.
- Less file flushing and faster performance when writing logs and committing to stores.
- More useful verbose output from some commands.
BUG FIXES:
- Fix inverted display of 'R' and 'M' during 'commit -v'.
PORTABILITY:
- Include a subset of ElementTree-1.2.20040618 to make installation easier.
- Fix time.localtime call to work with Python 2.3 (the minimum supported).
ENHANCEMENTS:
- First public release.
- Storage of local versions: init, add, remove, rm, info, log, diff, status, etc.