Changeset - ce5b7896d288
[Not reviewed]
default
0 1 0
Thomas De Schampheleire - 7 years ago 2019-03-28 21:49:07
thomas.de_schampheleire@nokia.com
git: fix handling of submodules that are not in the repo root (Issue #337)

GitChangeset.get_nodes() did not handle submodules correctly if they were
not located in the repository root. The file .gitmodules was searched in the
'tree' object being handled, which would be a subdirectory, while the real
.gitmodules file is in the root tree of the repository.

Instead of using 'name', the 'path' should be used.

This problem was noticed during indexing of such repositories.
1 file changed with 8 insertions and 7 deletions:
0 comments (0 inline, 0 general)
kallithea/lib/vcs/backends/git/changeset.py
Show inline comments
 
@@ -317,204 +317,205 @@ class GitChangeset(BaseChangeset):
 

	
 
        """
 
        self._get_filectx(path)
 
        from dulwich.walk import Walker
 
        include = [self.id]
 
        walker = Walker(self.repository._repo.object_store, include,
 
                        paths=[path], max_entries=1)
 
        return [self.repository.get_changeset(sha)
 
                for sha in (x.commit.id for x in walker)]
 

	
 
    def get_file_annotate(self, path):
 
        """
 
        Returns a generator of four element tuples with
 
            lineno, sha, changeset lazy loader and line
 

	
 
        TODO: This function now uses os underlying 'git' command which is
 
        generally not good. Should be replaced with algorithm iterating
 
        commits.
 
        """
 
        cmd = ['blame', '-l', '--root', '-r', self.id, '--', path]
 
        # -l     ==> outputs long shas (and we need all 40 characters)
 
        # --root ==> doesn't put '^' character for boundaries
 
        # -r sha ==> blames for the given revision
 
        so, se = self.repository.run_git_command(cmd)
 

	
 
        for i, blame_line in enumerate(so.split('\n')[:-1]):
 
            ln_no = i + 1
 
            sha, line = re.split(r' ', blame_line, 1)
 
            yield (ln_no, sha, lambda: self.repository.get_changeset(sha), line)
 

	
 
    def fill_archive(self, stream=None, kind='tgz', prefix=None,
 
                     subrepos=False):
 
        """
 
        Fills up given stream.
 

	
 
        :param stream: file like object.
 
        :param kind: one of following: ``zip``, ``tgz`` or ``tbz2``.
 
            Default: ``tgz``.
 
        :param prefix: name of root directory in archive.
 
            Default is repository name and changeset's raw_id joined with dash
 
            (``repo-tip.<KIND>``).
 
        :param subrepos: include subrepos in this archive.
 

	
 
        :raise ImproperArchiveTypeError: If given kind is wrong.
 
        :raise VcsError: If given stream is None
 

	
 
        """
 
        allowed_kinds = settings.ARCHIVE_SPECS.keys()
 
        if kind not in allowed_kinds:
 
            raise ImproperArchiveTypeError('Archive kind not supported use one'
 
                'of %s', allowed_kinds)
 

	
 
        if prefix is None:
 
            prefix = '%s-%s' % (self.repository.name, self.short_id)
 
        elif prefix.startswith('/'):
 
            raise VCSError("Prefix cannot start with leading slash")
 
        elif prefix.strip() == '':
 
            raise VCSError("Prefix cannot be empty")
 

	
 
        if kind == 'zip':
 
            frmt = 'zip'
 
        else:
 
            frmt = 'tar'
 
        _git_path = settings.GIT_EXECUTABLE_PATH
 
        cmd = '%s archive --format=%s --prefix=%s/ %s' % (_git_path,
 
                                                frmt, prefix, self.raw_id)
 
        if kind == 'tgz':
 
            cmd += ' | gzip -9'
 
        elif kind == 'tbz2':
 
            cmd += ' | bzip2 -9'
 

	
 
        if stream is None:
 
            raise VCSError('You need to pass in a valid stream for filling'
 
                           ' with archival data')
 
        popen = Popen(cmd, stdout=PIPE, stderr=PIPE, shell=True,
 
                      cwd=self.repository.path)
 

	
 
        buffer_size = 1024 * 8
 
        chunk = popen.stdout.read(buffer_size)
 
        while chunk:
 
            stream.write(chunk)
 
            chunk = popen.stdout.read(buffer_size)
 
        # Make sure all descriptors would be read
 
        popen.communicate()
 

	
 
    def get_nodes(self, path):
 
        if self._get_kind(path) != NodeKind.DIR:
 
            raise ChangesetError("Directory does not exist for revision %s at "
 
                " '%s'" % (self.revision, path))
 
        path = self._fix_path(path)
 
        id = self._get_id_for_path(path)
 
        tree = self.repository._repo[id]
 
        dirnodes = []
 
        filenodes = []
 
        als = self.repository.alias
 
        for name, stat, id in tree.iteritems():
 
            if path != '':
 
                obj_path = '/'.join((path, name))
 
            else:
 
                obj_path = name
 
            if objects.S_ISGITLINK(stat):
 
                cf = ConfigFile.from_file(BytesIO(self.repository._repo.get_object(tree['.gitmodules'][1]).data))
 
                url = cf.get(('submodule', name), 'url')
 
                dirnodes.append(SubModuleNode(name, url=url, changeset=id,
 
                root_tree = self.repository._repo[self._tree_id]
 
                cf = ConfigFile.from_file(BytesIO(self.repository._repo.get_object(root_tree['.gitmodules'][1]).data))
 
                url = cf.get(('submodule', obj_path), 'url')
 
                dirnodes.append(SubModuleNode(obj_path, url=url, changeset=id,
 
                                              alias=als))
 
                continue
 

	
 
            obj = self.repository._repo.get_object(id)
 
            if path != '':
 
                obj_path = '/'.join((path, name))
 
            else:
 
                obj_path = name
 
            if obj_path not in self._stat_modes:
 
                self._stat_modes[obj_path] = stat
 
            if isinstance(obj, objects.Tree):
 
                dirnodes.append(DirNode(obj_path, changeset=self))
 
            elif isinstance(obj, objects.Blob):
 
                filenodes.append(FileNode(obj_path, changeset=self, mode=stat))
 
            else:
 
                raise ChangesetError("Requested object should be Tree "
 
                                     "or Blob, is %r" % type(obj))
 
        nodes = dirnodes + filenodes
 
        for node in nodes:
 
            if node.path not in self.nodes:
 
                self.nodes[node.path] = node
 
        nodes.sort()
 
        return nodes
 

	
 
    def get_node(self, path):
 
        if isinstance(path, unicode):
 
            path = path.encode('utf-8')
 
        path = self._fix_path(path)
 
        if path not in self.nodes:
 
            try:
 
                id_ = self._get_id_for_path(path)
 
            except ChangesetError:
 
                raise NodeDoesNotExistError("Cannot find one of parents' "
 
                    "directories for a given path: %s" % path)
 

	
 
            _GL = lambda m: m and objects.S_ISGITLINK(m)
 
            if _GL(self._stat_modes.get(path)):
 
                tree = self.repository._repo[self._tree_id]
 
                cf = ConfigFile.from_file(BytesIO(self.repository._repo.get_object(tree['.gitmodules'][1]).data))
 
                url = cf.get(('submodule', path), 'url')
 
                node = SubModuleNode(path, url=url, changeset=id_,
 
                                     alias=self.repository.alias)
 
            else:
 
                obj = self.repository._repo.get_object(id_)
 

	
 
                if isinstance(obj, objects.Tree):
 
                    if path == '':
 
                        node = RootNode(changeset=self)
 
                    else:
 
                        node = DirNode(path, changeset=self)
 
                    node._tree = obj
 
                elif isinstance(obj, objects.Blob):
 
                    node = FileNode(path, changeset=self)
 
                    node._blob = obj
 
                else:
 
                    raise NodeDoesNotExistError("There is no file nor directory "
 
                        "at the given path '%s' at revision %s"
 
                        % (path, self.short_id))
 
            # cache node
 
            self.nodes[path] = node
 
        return self.nodes[path]
 

	
 
    @LazyProperty
 
    def affected_files(self):
 
        """
 
        Gets a fast accessible file changes for given changeset
 
        """
 
        added, modified, deleted = self._changes_cache
 
        return list(added.union(modified).union(deleted))
 

	
 
    @LazyProperty
 
    def _diff_name_status(self):
 
        output = []
 
        for parent in self.parents:
 
            cmd = ['diff', '--name-status', parent.raw_id, self.raw_id,
 
                   '--encoding=utf8']
 
            so, se = self.repository.run_git_command(cmd)
 
            output.append(so.strip())
 
        return '\n'.join(output)
 

	
 
    @LazyProperty
 
    def _changes_cache(self):
 
        added = set()
 
        modified = set()
 
        deleted = set()
 
        _r = self.repository._repo
 

	
 
        parents = self.parents
 
        if not self.parents:
 
            parents = [EmptyChangeset()]
 
        for parent in parents:
 
            if isinstance(parent, EmptyChangeset):
 
                oid = None
 
            else:
 
                oid = _r[parent.raw_id].tree
 
            changes = _r.object_store.tree_changes(oid, _r[self.raw_id].tree)
 
            for (oldpath, newpath), (_, _), (_, _) in changes:
 
                if newpath and oldpath:
 
                    modified.add(newpath)
 
                elif newpath and not oldpath:
 
                    added.add(newpath)
 
                elif not newpath and oldpath:
 
                    deleted.add(oldpath)
 
        return added, modified, deleted
0 comments (0 inline, 0 general)