~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2005 by Canonical Ltd

#

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

#

# This program is distributed in the hope that it will be useful,

# but WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

# GNU General Public License for more details.

#

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

"""Testament - a summary of a revision for signing.

A testament can be defined as "something that serves as tangible 

proof or evidence."  In bzr we use them to allow people to certify

particular revisions as authentic.  

The goal is that if two revisions are semantically equal, then they will

have a byte-for-byte equal testament.  We can define different versions of

"semantically equal" by using different testament classes; e.g. one that

includes or ignores file-ids.

We sign a testament rather than the revision XML itself for several reasons.

The most important is that the form in which the revision is stored

internally is designed for that purpose, and contains information which need

not be attested to by the signer.  For example the inventory contains the

last-changed revision for a file, but this is not necessarily something the

user cares to sign.

Having unnecessary fields signed makes the signatures brittle when the same

revision is stored in different branches or when the format is upgraded.

Handling upgrades is another motivation for using testaments separate from

the stored revision.  We would like to be able to compare a signature

generated from an old-format tree to newer tree, or vice versa.  This could

be done by comparing the revisions but that makes it unclear about exactly

what is being compared or not.

Different signing keys might indicate different levels of trust; we can in

the future extend this to allow signatures indicating not just that a

particular version is authentic but that it has other properties.

The signature can be applied to either the full testament or to just a

hash of it.

Testament format 1

~~~~~~~~~~~~~~~~~~

* timestamps are given as integers to avoid rounding errors

* parents given in lexicographical order

* indented-text form similar to log; intended to be human readable

* paths are given with forward slashes

* files are named using paths for ease of comparison/debugging

* the testament uses unix line-endings (\n)

"""

# XXX: At the moment, clients trust that the graph described in a weave

# is accurate, but that's not covered by the testament.  Perhaps the best

# fix is when verifying a revision to make sure that every file mentioned 

# in the revision has compatible ancestry links.

# TODO: perhaps write timestamp in a more readable form

# TODO: Perhaps these should just be different formats in which inventories/

# revisions can be serialized.

from copy import copy

from cStringIO import StringIO

import string

from sha import sha

from bzrlib.osutils import contains_whitespace, contains_linebreaks

class Testament(object):

    """Reduced summary of a revision.

    Testaments can be 

      - produced from a revision

      - written to a stream

      - loaded from a stream

      - compared to a revision

    """

    long_header = 'bazaar-ng testament version 1\n'

    short_header = 'bazaar-ng testament short form 1\n'

    @classmethod

    def from_revision(cls, repository, revision_id):

        """Produce a new testament from a historical revision"""

        rev = repository.get_revision(revision_id)

        inventory = repository.get_inventory(revision_id)

        return cls(rev, inventory)

    def __init__(self, rev, inventory):

        """Create a new testament for rev using inventory."""

        self.revision_id = str(rev.revision_id)

        self.committer = rev.committer

        self.timezone = rev.timezone or 0

        self.timestamp = rev.timestamp

        self.message = rev.message

        self.parent_ids = rev.parent_ids[:]

        self.inventory = inventory

        self.revprops = copy(rev.properties)

        assert not contains_whitespace(self.revision_id)

        assert not contains_linebreaks(self.committer)

    def as_text_lines(self):

        """Yield text form as a sequence of lines.

        The result is returned in utf-8, because it should be signed or

        hashed in that encoding.

        """

        r = []

        a = r.append

        a(self.long_header)

        a('revision-id: %s\n' % self.revision_id)

        a('committer: %s\n' % self.committer)

        a('timestamp: %d\n' % self.timestamp)

        a('timezone: %d\n' % self.timezone)

        # inventory length contains the root, which is not shown here

        a('parents:\n')

        for parent_id in sorted(self.parent_ids):

            assert not contains_whitespace(parent_id)

            a('  %s\n' % parent_id)

        a('message:\n')

        for l in self.message.splitlines():

            a('  %s\n' % l)

        a('inventory:\n')

        entries = self.inventory.iter_entries()

        entries.next()

        for path, ie in entries:

            a(self._entry_to_line(path, ie))

        r.extend(self._revprops_to_lines())

        if __debug__:

            for l in r:

                assert isinstance(l, basestring), \

                    '%r of type %s is not a plain string' % (l, type(l))

        return [line.encode('utf-8') for line in r]

    def _escape_path(self, path):

        assert not contains_linebreaks(path)

        return unicode(path.replace('\\', '/').replace(' ', '\ ')).encode('utf-8')

    def _entry_to_line(self, path, ie):

        """Turn an inventory entry into a testament line"""

        l = '  ' + str(ie.kind)

        l += ' ' + self._escape_path(path)

        assert not contains_whitespace(ie.file_id)

        l += ' ' + unicode(ie.file_id).encode('utf-8')

        if ie.kind == 'file':

            # TODO: avoid switching on kind

            assert ie.text_sha1

            l += ' ' + ie.text_sha1

        elif ie.kind == 'symlink':

            assert ie.symlink_target

            l += ' ' + self._escape_path(ie.symlink_target)

        l += '\n'

        return l.decode('utf-8')

    def as_text(self):

        return ''.join(self.as_text_lines())

    def as_short_text(self):

        """Return short digest-based testament."""

        return (self.short_header + 

                'revision-id: %s\n'

                'sha1: %s\n'

                % (self.revision_id, self.as_sha1()))

    def _revprops_to_lines(self):

        """Pack up revision properties."""

        if not self.revprops:

            return []

        r = ['properties:\n']

        for name, value in sorted(self.revprops.items()):

            assert isinstance(name, str)

            assert not contains_whitespace(name)

            r.append('  %s:\n' % name)

            for line in value.splitlines():

                if not isinstance(line, str):

                    line = line.encode('utf-8')

                r.append('    %s\n' % line)

        return r

    def as_sha1(self):

        s = sha()

        map(s.update, self.as_text_lines())

        return s.hexdigest()

class StrictTestament(Testament):

    """This testament format is for use as a checksum in changesets"""

    long_header = 'bazaar-ng testament version 2.1\n'

    short_header = 'bazaar-ng testament short form 2.1\n'

    def _entry_to_line(self, path, ie):

        l = Testament._entry_to_line(self, path, ie)[:-1]

        l += ' ' + ie.revision.decode('utf-8')

        l += {True: ' yes\n', False: ' no\n'}[ie.executable]

        return l