# obsolete.py - obsolete markers handling

#

# Copyright 2012 Pierre-Yves David <pierre-yves.david@ens-lyon.org>

#                Logilab SA        <contact@logilab.fr>

#

# This software may be used and distributed according to the terms of the

# GNU General Public License version 2 or any later version.

"""Obsolete marker handling

An obsolete marker maps an old changeset to a list of new

changesets. If the list of new changesets is empty, the old changeset

is said to be "killed". Otherwise, the old changeset is being

"replaced" by the new changesets.

Obsolete markers can be used to record and distribute changeset graph

transformations performed by history rewrite operations, and help

building new tools to reconcile conflicting rewrite actions. To

facilitate conflict resolution, markers include various annotations

besides old and news changeset identifiers, such as creation date or

author name.

The old obsoleted changeset is called a "predecessor" and possible

replacements are called "successors". Markers that used changeset X as

a predecessor are called "successor markers of X" because they hold

information about the successors of X. Markers that use changeset Y as

a successors are call "predecessor markers of Y" because they hold

information about the predecessors of Y.

Examples:

- When changeset A is replaced by changeset A', one marker is stored:

    (A, (A',))

- When changesets A and B are folded into a new changeset C, two markers are

  stored:

    (A, (C,)) and (B, (C,))

- When changeset A is simply "pruned" from the graph, a marker is created:

    (A, ())

- When changeset A is split into B and C, a single marker is used:

    (A, (B, C))

  We use a single marker to distinguish the "split" case from the "divergence"

  case. If two independent operations rewrite the same changeset A in to A' and

  A'', we have an error case: divergent rewriting. We can detect it because

  two markers will be created independently:

  (A, (B,)) and (A, (C,))

Format

------

Markers are stored in an append-only file stored in

'.hg/store/obsstore'.

The file starts with a version header:

- 1 unsigned byte: version number, starting at zero.

The header is followed by the markers. Marker format depend of the version. See

comment associated with each format for details.

"""

import binascii

import struct

import weakref

from .i18n import _

from .pycompat import getattr

from .node import (

    bin,

    hex,

)

from . import (

    encoding,

    error,

    obsutil,

    phases,

    policy,

    pycompat,

    util,

)

from .utils import (

    dateutil,

    hashutil,

)

parsers = policy.importmod('parsers')

_pack = struct.pack

_unpack = struct.unpack

_calcsize = struct.calcsize

propertycache = util.propertycache

# Options for obsolescence

createmarkersopt = b'createmarkers'

allowunstableopt = b'allowunstable'

allowdivergenceopt = b'allowdivergence'

exchangeopt = b'exchange'

def _getoptionvalue(repo, option):

    """Returns True if the given repository has the given obsolete option

    enabled.

    """

    configkey = b'evolution.%s' % option

    newconfig = repo.ui.configbool(b'experimental', configkey)

    # Return the value only if defined

    if newconfig is not None:

        return newconfig

    # Fallback on generic option

    try:

        return repo.ui.configbool(b'experimental', b'evolution')

    except (error.ConfigError, AttributeError):

        # Fallback on old-fashion config

        # inconsistent config: experimental.evolution

        result = set(repo.ui.configlist(b'experimental', b'evolution'))

        if b'all' in result:

            return True

        # Temporary hack for next check

        newconfig = repo.ui.config(b'experimental', b'evolution.createmarkers')

        if newconfig:

            result.add(b'createmarkers')

        return option in result

def getoptions(repo):

    """Returns dicts showing state of obsolescence features."""

    createmarkersvalue = _getoptionvalue(repo, createmarkersopt)

    if createmarkersvalue:

        unstablevalue = _getoptionvalue(repo, allowunstableopt)

        divergencevalue = _getoptionvalue(repo, allowdivergenceopt)

        exchangevalue = _getoptionvalue(repo, exchangeopt)

    else:

        # if we cannot create obsolescence markers, we shouldn't exchange them

        # or perform operations that lead to instability or divergence

        unstablevalue = False

        divergencevalue = False

        exchangevalue = False

    return {

        createmarkersopt: createmarkersvalue,

        allowunstableopt: unstablevalue,

        allowdivergenceopt: divergencevalue,

        exchangeopt: exchangevalue,

    }

def isenabled(repo, option):

    """Returns True if the given repository has the given obsolete option

    enabled.

    """

    return getoptions(repo)[option]

# Creating aliases for marker flags because evolve extension looks for

# bumpedfix in obsolete.py

bumpedfix = obsutil.bumpedfix

usingsha256 = obsutil.usingsha256

## Parsing and writing of version "0"

#

# The header is followed by the markers. Each marker is made of:

#

# - 1 uint8 : number of new changesets "N", can be zero.

#

# - 1 uint32: metadata size "M" in bytes.

#

# - 1 byte: a bit field. It is reserved for flags used in common

#   obsolete marker operations, to avoid repeated decoding of metadata

#   entries.

#

# - 20 bytes: obsoleted changeset identifier.

#

# - N*20 bytes: new changesets identifiers.

#

# - M bytes: metadata as a sequence of nul-terminated strings. Each

#   string contains a key and a value, separated by a colon ':', without

#   additional encoding. Keys cannot contain '\0' or ':' and values

#   cannot contain '\0'.

_fm0version = 0

_fm0fixed = b'>BIB20s'

_fm0node = b'20s'

_fm0fsize = _calcsize(_fm0fixed)

_fm0fnodesize = _calcsize(_fm0node)

def _fm0readmarkers(data, off, stop):

    # Loop on markers

    while off < stop:

        # read fixed part

        cur = data[off : off + _fm0fsize]

        off += _fm0fsize

        numsuc, mdsize, flags, pre = _unpack(_fm0fixed, cur)

        # read replacement

        sucs = ()

        if numsuc:

            s = _fm0fnodesize * numsuc

            cur = data[off : off + s]

            sucs = _unpack(_fm0node * numsuc, cur)

            off += s

        # read metadata

        # (metadata will be decoded on demand)

        metadata = data[off : off + mdsize]

        if len(metadata) != mdsize:

            raise error.Abort(

                _(

                    b'parsing obsolete marker: metadata is too '

                    b'short, %d bytes expected, got %d'

                )

                % (mdsize, len(metadata))

            )

        off += mdsize

        metadata = _fm0decodemeta(metadata)

        try:

            when, offset = metadata.pop(b'date', b'0 0').split(b' ')

            date = float(when), int(offset)

        except ValueError:

            date = (0.0, 0)

        parents = None

        if b'p2' in metadata:

            parents = (metadata.pop(b'p1', None), metadata.pop(b'p2', None))

        elif b'p1' in metadata:

            parents = (metadata.pop(b'p1', None),)

        elif b'p0' in metadata:

            parents = ()

        if parents is not None:

            try:

                parents = tuple(bin(p) for p in parents)

                # if parent content is not a nodeid, drop the data

                for p in parents:

                    if len(p) != 20:

                        parents = None

                        break

            except binascii.Error:

                # if content cannot be translated to nodeid drop the data.

                parents = None

        metadata = tuple(sorted(metadata.items()))

        yield (pre, sucs, flags, metadata, date, parents)

def _fm0encodeonemarker(marker):

    pre, sucs, flags, metadata, date, parents = marker

    if flags & usingsha256:

        raise error.Abort(_(b'cannot handle sha256 with old obsstore format'))

    metadata = dict(metadata)

    time, tz = date

    metadata[b'date'] = b'%r %i' % (time, tz)

    if parents is not None:

        if not parents:

            # mark that we explicitly recorded no parents

            metadata[b'p0'] = b''

        for i, p in enumerate(parents, 1):

            metadata[b'p%i' % i] = hex(p)

    metadata = _fm0encodemeta(metadata)

    numsuc = len(sucs)

    format = _fm0fixed + (_fm0node * numsuc)

    data = [numsuc, len(metadata), flags, pre]

    data.extend(sucs)

    return _pack(format, *data) + metadata

def _fm0encodemeta(meta):

    """Return encoded metadata string to string mapping.

    Assume no ':' in key and no '\0' in both key and value."""

    for key, value in meta.items():

        if b':' in key or b'\0' in key:

            raise ValueError(b"':' and '\0' are forbidden in metadata key'")

        if b'\0' in value:

            raise ValueError(b"':' is forbidden in metadata value'")

    return b'\0'.join([b'%s:%s' % (k, meta[k]) for k in sorted(meta)])

def _fm0decodemeta(data):

    """Return string to string dictionary from encoded version."""

    d = {}

    for l in data.split(b'\0'):

        if l:

            key, value = l.split(b':', 1)

            d[key] = value

    return d

## Parsing and writing of version "1"

#

# The header is followed by the markers. Each marker is made of:

#

# - uint32: total size of the marker (including this field)

#

# - float64: date in seconds since epoch

#

# - int16: timezone offset in minutes

#

# - uint16: a bit field. It is reserved for flags used in common

#   obsolete marker operations, to avoid repeated decoding of metadata

#   entries.

#

# - uint8: number of successors "N", can be zero.

#

# - uint8: number of parents "P", can be zero.

#

#     0: parents data stored but no parent,

#     1: one parent stored,

#     2: two parents stored,

#     3: no parent data stored

#

# - uint8: number of metadata entries M

#

# - 20 or 32 bytes: predecessor changeset identifier.

#

# - N*(20 or 32) bytes: successors changesets identifiers.

#

# - P*(20 or 32) bytes: parents of the predecessors changesets.

#

# - M*(uint8, uint8): size of all metadata entries (key and value)

#

# - remaining bytes: the metadata, each (key, value) pair after the other.

_fm1version = 1

_fm1fixed = b'>IdhHBBB'

_fm1nodesha1 = b'20s'

_fm1nodesha256 = b'32s'

_fm1nodesha1size = _calcsize(_fm1nodesha1)

_fm1nodesha256size = _calcsize(_fm1nodesha256)

_fm1fsize = _calcsize(_fm1fixed)

_fm1parentnone = 3

_fm1metapair = b'BB'

_fm1metapairsize = _calcsize(_fm1metapair)

def _fm1purereadmarkers(data, off, stop):

    # make some global constants local for performance

    noneflag = _fm1parentnone

    sha2flag = usingsha256

    sha1size = _fm1nodesha1size

    sha2size = _fm1nodesha256size

    sha1fmt = _fm1nodesha1

    sha2fmt = _fm1nodesha256

    metasize = _fm1metapairsize

    metafmt = _fm1metapair

    fsize = _fm1fsize

    unpack = _unpack

    # Loop on markers

    ufixed = struct.Struct(_fm1fixed).unpack

    while off < stop:

        # read fixed part

        o1 = off + fsize

        t, secs, tz, flags, numsuc, numpar, nummeta = ufixed(data[off:o1])

        if flags & sha2flag:

            nodefmt = sha2fmt

            nodesize = sha2size

        else:

            nodefmt = sha1fmt

            nodesize = sha1size

        (prec,) = unpack(nodefmt, data[o1 : o1 + nodesize])

        o1 += nodesize

        # read 0 or more successors

        if numsuc == 1:

            o2 = o1 + nodesize

            sucs = (data[o1:o2],)

        else:

            o2 = o1 + nodesize * numsuc

            sucs = unpack(nodefmt * numsuc, data[o1:o2])

        # read parents

        if numpar == noneflag:

            o3 = o2

            parents = None

        elif numpar == 1:

            o3 = o2 + nodesize

            parents = (data[o2:o3],)

        else:

            o3 = o2 + nodesize * numpar

            parents = unpack(nodefmt * numpar, data[o2:o3])

        # read metadata

        off = o3 + metasize * nummeta

        metapairsize = unpack(b'>' + (metafmt * nummeta), data[o3:off])

        metadata = []

        for idx in range(0, len(metapairsize), 2):

            o1 = off + metapairsize[idx]

            o2 = o1 + metapairsize[idx + 1]