New patches: [bzr API: tree inventory Lalo Martins **20060328092456 The bzr API changed again; now the Branch object doesn't provide a get_inventory() method anymore, since conceptually, branches don't have inventories (branches have trees, which have inventories). ] { hunk ./vcpx/bzr.py 134 - inv = self._b.get_inventory(self._b.last_revision()) - diff = compare_trees(self._b.revision_tree(last_revision), - self._b.working_tree()) + basis_tree = self._b.basis_tree() + inv = basis_tree.inventory + diff = compare_trees(basis_tree, self._b.working_tree()) } [bzrlib 0.8 stage 1 Lalo Martins **20060330081514 Partial upgrade of bzr.py for the bzrlib 0.8 api. Missing only _applyChangeset() which is causing me quite a headache. ] { hunk ./vcpx/bzr.py 6 +# Lalo Martins hunk ./vcpx/bzr.py 16 +import os +from workdir import WorkingDir hunk ./vcpx/bzr.py 20 +from bzrlib.bzrdir import BzrDir hunk ./vcpx/bzr.py 23 +from bzrlib import errors hunk ./vcpx/bzr.py 26 + def __init__(self, repository): + WorkingDir.__init__(self, repository) + # TODO: check if there is a "repository" in the configuration, + # and use it as a bzr repository + self._branch = None + self._working_tree = None + try: + self._bzrdir = BzrDir.open(self.basedir) + except errors.NotBranchError: + self._bzrdir = None + else: + self.__open_branch() + + def __open_branch(self): + if self._branch is not None: + print repr(self._branch), 'already opened' + return + if self._bzrdir is None: + raise RuntimeError("Can't open branch if bzrdir doesn't exist") + self._branch = self._bzrdir.open_branch() + self._working_tree = self._bzrdir.open_workingtree() + + ############################# hunk ./vcpx/bzr.py 57 - r = parent.get_revision(revision) + r = parent.repository.get_revision(revision) hunk ./vcpx/bzr.py 93 - from bzrlib import fetch + self.__open_branch() hunk ./vcpx/bzr.py 95 - parent = Branch.open(self.repository.repository) - repo = self._getRepo() + parent = BzrDir.open(self.repository.repository).open_branch() hunk ./vcpx/bzr.py 97 - revisions = repo.missing_revisions(parent) - fetch.greedy_fetch(repo, parent) + revisions = self._branch.missing_revisions(parent) + self._branch.fetch(parent) hunk ./vcpx/bzr.py 107 + raise NotImplemented("this method wasn't upgraded to " + "bzrlib 0.8 yet, sorry") + hunk ./vcpx/bzr.py 112 - repo = self._getRepo() - parent = Branch.open(self.repository.repository) + self.__open_branch() hunk ./vcpx/bzr.py 114 - oldrevno = repo.revno() + parent = BzrDir.open(self.repository.repository).open_branch() + + oldrevno = self._branch.revno() hunk ./vcpx/bzr.py 119 - repo.append_revision(changeset.revision) + self._branch.append_revision(changeset.revision) hunk ./vcpx/bzr.py 131 - from bzrlib.clone import copy_branch - - parent = Branch.open(self.repository.repository) + parent = BzrDir.open(self.repository.repository) + parent_branch = parent.open_branch() hunk ./vcpx/bzr.py 135 - revid = parent.get_rev_id(1) + revid = parent_branch.get_rev_id(1) hunk ./vcpx/bzr.py 143 - self._b = copy_branch(parent, self.basedir, revid) + self._bzrdir = parent.clone(self.basedir, revid) hunk ./vcpx/bzr.py 145 - return self._changesetFromRevision(parent, revid) + return self._changesetFromRevision(parent_branch, revid) hunk ./vcpx/bzr.py 147 + ################################# hunk ./vcpx/bzr.py 159 - last_revision = self._b.last_revision() + self.__open_branch() + + last_revision = self._branch.last_revision() hunk ./vcpx/bzr.py 167 - basis_tree = self._b.basis_tree() + basis_tree = self._branch.basis_tree() hunk ./vcpx/bzr.py 169 - diff = compare_trees(basis_tree, self._b.working_tree()) + diff = compare_trees(basis_tree, self._working_tree) hunk ./vcpx/bzr.py 191 - self._b.working_tree().add(new_entries) + self._working_tree.add(new_entries) hunk ./vcpx/bzr.py 221 - self._b.working_tree().commit(logmessage, committer=author, + self._bzrdir.open_workingtree().commit(logmessage, committer=author, hunk ./vcpx/bzr.py 230 - self._b.working_tree().remove(entries) + self._working_tree.remove(entries) hunk ./vcpx/bzr.py 243 - self._b.working_tree().rename_one(oldentry, newentry) + self._working_tree.rename_one(oldentry, newentry) hunk ./vcpx/bzr.py 254 - if not exists(join(self.basedir, self.repository.METADIR)): + if self._bzrdir is None: hunk ./vcpx/bzr.py 277 - self._b = Branch.initialize(self.basedir) + self._bzrdir = BzrDir.create(self.basedir) + self._branch = self._bzrdir.create_branch() + self._working_tree = self._bzrdir.create_workingtree() hunk ./vcpx/bzr.py 281 - self._b = Branch.open(self.basedir) - - def _getRepo(self): - try: - return self._b - except AttributeError: - self._b = Branch.open(self.basedir) - return self._b + self.__open_branch() } Context: [Update hglib backend to work with recent versions of Mercurial. Travis Cross **20060224200653 Added required parameters to the calls to pull and clone. ] [Fix to hglib: don't try to .split() date as it is already a tuple. Travis Cross **20060224200525] [In hglib, check if a file is in the manifest before trying to rename it. Travis Cross **20060224200005] [Separate option name from its value lele@metapensiero.it**20060221150654] [Remove spurious and dead code lele@metapensiero.it**20060221110359 I'm not able to trace the origin of this code, reasonably that loop was in a support method and then moved inside __init__(). Good catch, Evgeny! ] [Suggest #tailor IRC channel, not #darcs lele@metapensiero.it**20060216205756] [Disallow encodings other than utf-8 for mercurial Brendan Cully **20060216201024] [hglib doesn't accept unicode strings for path names, so encode them Brendan Cully **20060216200909] [Don't force the LANG on svn commit lele@nautilus.homeip.net**20060214163944 Tailor used LANG=C to be able to catch the revision number generated by commit. It now assumes the revision number is the integer value printed out by svn in the last line of commit output. ] [Clever change to stop misreading the boolean expression as a keyword argument lele@nautilus.homeip.net**20060125235421] [Fix handling of tla archives with "untagged-source unrecognized" set Cameron Patrick **20060212173351 If a tla archive has the "untagged-source unrecognized" option in its tagging-method file, tla will complain if there are any foreign files present (e.g. the _darcs directory containing darcs metadata). The tla backend was supposed to fix this but the code there didn't work properly, as it was too fussy about tla exit codes and the output formats of `tla tree-lint`. This patch makes it work again (tested with tla 1.3.3). ] [More informative error messages lele@metapensiero.it**20060209194125] [Fix to #30: handle CVS target, creating repository and working dir lele@metapensiero.it**20060209193342] [Use "initialize+pull" instead of "get" when the directory already exist lele@metapensiero.it**20060209193158] [Added a comment at top to clarify the strange markup lele@metapensiero.it**20060203192712] [Fix typo lele@metapensiero.it**20060203192642] [Convert the delay into a float lele@metapensiero.it**20060128112524 This fixes #34. ] [Test also the default case for look-for-adds lele@metapensiero.it**20060124172039] [Allow 'look-for-adds' for darcs repositories lele@metapensiero.it**20060124171536 I'm little bit worried that this may hide real bugs, but here it is. ] [When collapsing two changesets into one, make sure the tags from the second go into the combined one. Fixes ticket #26. tailor23494@aaronkaplan.info**20060120225053] [Do not stop on GetUpstreamChangesetsFailure lele@metapensiero.it**20060118105809 When operating on several projects, do not stop on this kind of error, that most often is related to overloaded CVS servers, but keep going with the other projects. ] [There's no "string" builtin type, use "basestring" instead. lele@metapensiero.it**20060118091638] [Annotate the characters that tailor washes out of Subversion XML logs lele@metapensiero.it**20060118091528 when filter-badchars is set to True. ] [Fix cut-n-paste error lele@metapensiero.it**20051228174346 This most probably caused some breakage in projects using distinct shared directories for source and target repositories. ] [M-x whitespace-cleanup lele@metapensiero.it**20051227134851] [Do not add entries inside a moved/copied directory on bzr target lele@metapensiero.it**20051227134834 The fix to #24 and indirectly #23 triggered (actually, the test did) a similar issue solved for Subversion with [840] and [841]. This patch applies a similar solution for BazaarNG, to prevent addition within a renamed/copied directory. Specifically, this may happen for example in a changeset coming from Svn that 1. copies a directory 2. replaces (a Subversion "R" event) an item in the target with something else ] [Reset the notion of external copies at each changeset lele@metapensiero.it**20051227122808 Changesets [840] and [841] introduced a bug in the Subversion log parser, outlined by #24. This fix the case by resetting the list of copied paths at each changeset. ] [Avoid execution of too long command lines lele@metapensiero.it**20051226181934 Now the mechanism that executes external commands respects a maximum size for the command line, eventually executing the external command multiple times operating on a slice of the original arguments at a time. ] [Do not use bzr's smart_add_tree as it ignores entries marienz@gentoo.org**20051226094327 Also, do not add rename targets as they are already considered by bzr. ] [Parametrize the temporary directory used by tests lele@metapensiero.it**20051224190640] [Simplify rst2html command line lele@metapensiero.it**20051224190005 Recent docutils automatically embeds the right stylesheet. ] [Explicitly deprecate the current svndump backend lele@metapensiero.it**20051224185919] [New option ignore-externals on svn source backend lele@metapensiero.it**20051224185547 This **changes** the default behaviour of Tailor svn source backend: now it uses `--ignore-externals` to checkouts and updates. You can override the default putting "ignore-externals = False" in the svn repository section. ] [TAG Version 0.9.20 lele@metapensiero.it**20051223153837] Patch bundle hash: ab68a0bab18368b4cfa98ce7bb0b4b0ed56dd8a5