view mercurial/sshpeer.py @ 48687:f8f2ecdde4b5

branchmap: skip obsolete revisions while computing heads

It's time to make this part of core Mercurial obsolescence-aware.

Not considering obsolete revisions when computing heads is clearly what
Mercurial should do. But there are a couple of small issues:

- Let's say tip of the repo is obsolete. There are two ways of finding tiprev
  for branchcache (both are in use): looking at input data for update() and
  looking at computed heads after update(). Previously, repo tip would be
  tiprev of the branchcache. With this patch, an obsolete revision can no
  longer be tiprev. And depending on what way we use for finding tiprev (input
  data vs computed heads) we'll get a different result. This is relevant when
  recomputing cache key from cache contents, and may lead to updating cache for
  obsolete revisions multiple times (not from scratch, because it still would
  be considered valid for a subset of revisions in the repo).

- If all commits on a branch are obsolete, the branchcache will include that
  branch, but the list of heads will be empty (that's why there's now `if not
  heads` when recomputing tiprev/tipnode from cache contents). Having an entry
  for every branch is currently required for notify extension (and
  test-notify.t to pass), because notify doesn't handle revsets in its
  subscription config very well and will throw an error if e.g. a branch
  doesn't exist.

- Cloning static HTTP repos may try to stat() a non-existent obsstore file. The
  issue is that we now care about obsolescence during clone, but statichttpvfs
  doesn't implement a stat method, so a regular vfs.stat() is used, and it
  assumes that file is local and calls os.stat().  During a clone, we're trying
  to stat() .hg/store/obsstore, but in static HTTP case we provide a literal
  URL to the obsstore file on the remote as if it were a local file path. On
  windows it actually results in a failure in test-static-http.t.

The first issue is going to be addressed in a series dedicated to making sure
branchcache is properly and timely written on disk (it wasn't perfect even
before this patch, but there aren't enough tests to demonstrate that). The
second issue will be addressed in a future patch for notify extension that will
make it not raise an exception if a branch doesn't exist. And the third one was
partially addressed in the previous patch in this series and will be properly
fixed in a future patch when this series is accepted.

filteredhash() grows a keyword argument to make sure that branchcache is also
invalidated when there are new obsolete revisions in its repo view. This way
the on-disk cache format is unchanged and compatible between versions (although
it will obviously be recomputed when switching versions before/after this patch
and the repo has obsolete revisions).

There's one test that uses plain `hg up` without arguments while updated to a
pruned commit. To make this test pass, simply return current working directory
parent. Later in this series this code will be replaced by what prune command
does: updating to the closest non-obsolete ancestor.

Test changes:

test-branch-change.t: update branch head and cache update message. The head of
default listed in hg heads is changed because revision 2 was rewritten as 7,
and 1 is the closest ancestor on the same branch, so it's the head of default
now.

The cache invalidation message appears now because of the cache hash change,
since we're now accounting for obsolete revisions. Here's some context:

"served.hidden" repo filter means everything is visible (no filtered
revisions), so before this series branch2-served.hidden file would not contain
any cache hash, only revnum and node. Now it also has a hash when there are
obsolete changesets in the repo. The command that the message appears for is
changing branch of 5 and 6, which are now obsolete, so the cache hash changes.

In general, when cache is simply out-of-date, it can be updated using the old
version as a base. But if cache hash differs, then the cache for that
particular repo filter is recomputed (at least with the current
implementation). This is what happens here.

test-obsmarker-template.t: the pull reports 2 heads changed, but after that the
repo correctly sees only 1. The new message could be better, but it's still an
improvement over the previous one where hg pull suggested merging with an
obsolete revision.

test-obsolete.t: we can see these revisions in hg log --hidden, but they
shouldn't be considered heads even with --hidden.

test-rebase-obsolete{,2}.t: there were new heads created previously after
making new orphan changesets, but they weren't detected. Now we are properly
detecting and reporting them.

test-rebase-obsolete4.t: there's only one head now because the other head is
pruned and was falsely reported before.

test-static-http.t: add obsstore to the list of requested files. This file
doesn't exist on the remotes, but clients want it anyway (they get 404). This
is fine, because there are other nonexistent files that clients request, like
.hg/bookmarks or .hg/cache/tags2-served.

Differential Revision: https://phab.mercurial-scm.org/D12097

# sshpeer.py - ssh repository proxy class for mercurial
#
# Copyright 2005, 2006 Olivia Mackall <[e-mail]>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

from __future__ import absolute_import

import re
import uuid

from .i18n import _
from .pycompat import getattr
from . import (
    error,
    pycompat,
    util,
    wireprototypes,
    wireprotov1peer,
    wireprotov1server,
)
from .utils import (
    procutil,
    stringutil,
    urlutil,
)


def _serverquote(s):
    """quote a string for the remote shell ... which we assume is sh"""
    if not s:
        return s
    if re.match(b'[a-zA-Z0-9@%_+=:,./-]*$', s):
        return s
    return b"'%s'" % s.replace(b"'", b"'\\''")


def _forwardoutput(ui, pipe, warn=False):
    """display all data currently available on pipe as remote output.

    This is non blocking."""
    if pipe and not pipe.closed:
        s = procutil.readpipe(pipe)
        if s:
            display = ui.warn if warn else ui.status
            for l in s.splitlines():
                display(_(b"remote: "), l, b'\n')


class doublepipe(object):
    """Operate a side-channel pipe in addition of a main one

    The side-channel pipe contains server output to be forwarded to the user
    input. The double pipe will behave as the "main" pipe, but will ensure the
    content of the "side" pipe is properly processed while we wait for blocking
    call on the "main" pipe.

    If large amounts of data are read from "main", the forward will cease after
    the first bytes start to appear. This simplifies the implementation
    without affecting actual output of sshpeer too much as we rarely issue
    large read for data not yet emitted by the server.

    The main pipe is expected to be a 'bufferedinputpipe' from the util module
    that handle all the os specific bits. This class lives in this module
    because it focus on behavior specific to the ssh protocol."""

    def __init__(self, ui, main, side):
        self._ui = ui
        self._main = main
        self._side = side

    def _wait(self):
        """wait until some data are available on main or side

        return a pair of boolean (ismainready, issideready)

        (This will only wait for data if the setup is supported by `util.poll`)
        """
        if (
            isinstance(self._main, util.bufferedinputpipe)
            and self._main.hasbuffer
        ):
            # Main has data. Assume side is worth poking at.
            return True, True

        fds = [self._main.fileno(), self._side.fileno()]
        try:
            act = util.poll(fds)
        except NotImplementedError:
            # non supported yet case, assume all have data.
            act = fds
        return (self._main.fileno() in act, self._side.fileno() in act)

    def write(self, data):
        return self._call(b'write', data)

    def read(self, size):
        r = self._call(b'read', size)
        if size != 0 and not r:
            # We've observed a condition that indicates the
            # stdout closed unexpectedly. Check stderr one
            # more time and snag anything that's there before
            # letting anyone know the main part of the pipe
            # closed prematurely.
            _forwardoutput(self._ui, self._side)
        return r

    def unbufferedread(self, size):
        r = self._call(b'unbufferedread', size)
        if size != 0 and not r:
            # We've observed a condition that indicates the
            # stdout closed unexpectedly. Check stderr one
            # more time and snag anything that's there before
            # letting anyone know the main part of the pipe
            # closed prematurely.
            _forwardoutput(self._ui, self._side)
        return r

    def readline(self):
        return self._call(b'readline')

    def _call(self, methname, data=None):
        """call <methname> on "main", forward output of "side" while blocking"""
        # data can be '' or 0
        if (data is not None and not data) or self._main.closed:
            _forwardoutput(self._ui, self._side)
            return b''
        while True:
            mainready, sideready = self._wait()
            if sideready:
                _forwardoutput(self._ui, self._side)
            if mainready:
                meth = getattr(self._main, methname)
                if data is None:
                    return meth()
                else:
                    return meth(data)

    def close(self):
        return self._main.close()

    @property
    def closed(self):
        return self._main.closed

    def flush(self):
        return self._main.flush()


def _cleanuppipes(ui, pipei, pipeo, pipee, warn):
    """Clean up pipes used by an SSH connection."""
    didsomething = False
    if pipeo and not pipeo.closed:
        didsomething = True
        pipeo.close()
    if pipei and not pipei.closed:
        didsomething = True
        pipei.close()

    if pipee and not pipee.closed:
        didsomething = True
        # Try to read from the err descriptor until EOF.
        try:
            for l in pipee:
                ui.status(_(b'remote: '), l)
        except (IOError, ValueError):
            pass

        pipee.close()

    if didsomething and warn is not None:
        # Encourage explicit close of sshpeers. Closing via __del__ is
        # not very predictable when exceptions are thrown, which has led
        # to deadlocks due to a peer get gc'ed in a fork
        # We add our own stack trace, because the stacktrace when called
        # from __del__ is useless.
        ui.develwarn(b'missing close on SSH connection created at:\n%s' % warn)


def _makeconnection(ui, sshcmd, args, remotecmd, path, sshenv=None):
    """Create an SSH connection to a server.

    Returns a tuple of (process, stdin, stdout, stderr) for the
    spawned process.
    """
    cmd = b'%s %s %s' % (
        sshcmd,
        args,
        procutil.shellquote(
            b'%s -R %s serve --stdio'
            % (_serverquote(remotecmd), _serverquote(path))
        ),
    )

    ui.debug(b'running %s\n' % cmd)

    # no buffer allow the use of 'select'
    # feel free to remove buffering and select usage when we ultimately
    # move to threading.
    stdin, stdout, stderr, proc = procutil.popen4(cmd, bufsize=0, env=sshenv)

    return proc, stdin, stdout, stderr


def _clientcapabilities():
    """Return list of capabilities of this client.

    Returns a list of capabilities that are supported by this client.
    """
    protoparams = {b'partial-pull'}
    comps = [
        e.wireprotosupport().name
        for e in util.compengines.supportedwireengines(util.CLIENTROLE)
    ]
    protoparams.add(b'comp=%s' % b','.join(comps))
    return protoparams


def _performhandshake(ui, stdin, stdout, stderr):
    def badresponse():
        # Flush any output on stderr. In general, the stderr contains errors
        # from the remote (ssh errors, some hg errors), and status indications
        # (like "adding changes"), with no current way to tell them apart.
        # Here we failed so early that it's almost certainly only errors, so
        # use warn=True so -q doesn't hide them.
        _forwardoutput(ui, stderr, warn=True)

        msg = _(b'no suitable response from remote hg')
        hint = ui.config(b'ui', b'ssherrorhint')
        raise error.RepoError(msg, hint=hint)

    # The handshake consists of sending wire protocol commands in reverse
    # order of protocol implementation and then sniffing for a response
    # to one of them.
    #
    # Those commands (from oldest to newest) are:
    #
    # ``between``
    #   Asks for the set of revisions between a pair of revisions. Command
    #   present in all Mercurial server implementations.
    #
    # ``hello``
    #   Instructs the server to advertise its capabilities. Introduced in
    #   Mercurial 0.9.1.
    #
    # ``upgrade``
    #   Requests upgrade from default transport protocol version 1 to
    #   a newer version. Introduced in Mercurial 4.6 as an experimental
    #   feature.
    #
    # The ``between`` command is issued with a request for the null
    # range. If the remote is a Mercurial server, this request will
    # generate a specific response: ``1\n\n``. This represents the
    # wire protocol encoded value for ``\n``. We look for ``1\n\n``
    # in the output stream and know this is the response to ``between``
    # and we're at the end of our handshake reply.
    #
    # The response to the ``hello`` command will be a line with the
    # length of the value returned by that command followed by that
    # value. If the server doesn't support ``hello`` (which should be
    # rare), that line will be ``0\n``. Otherwise, the value will contain
    # RFC 822 like lines. Of these, the ``capabilities:`` line contains
    # the capabilities of the server.
    #
    # The ``upgrade`` command isn't really a command in the traditional
    # sense of version 1 of the transport because it isn't using the
    # proper mechanism for formatting insteads: instead, it just encodes
    # arguments on the line, delimited by spaces.
    #
    # The ``upgrade`` line looks like ``upgrade <token> <capabilities>``.
    # If the server doesn't support protocol upgrades, it will reply to
    # this line with ``0\n``. Otherwise, it emits an
    # ``upgraded <token> <protocol>`` line to both stdout and stderr.
    # Content immediately following this line describes additional
    # protocol and server state.
    #
    # In addition to the responses to our command requests, the server
    # may emit "banner" output on stdout. SSH servers are allowed to
    # print messages to stdout on login. Issuing commands on connection
    # allows us to flush this banner output from the server by scanning
    # for output to our well-known ``between`` command. Of course, if
    # the banner contains ``1\n\n``, this will throw off our detection.

    requestlog = ui.configbool(b'devel', b'debug.peer-request')

    # Generate a random token to help identify responses to version 2
    # upgrade request.
    token = pycompat.sysbytes(str(uuid.uuid4()))

    try:
        pairsarg = b'%s-%s' % (b'0' * 40, b'0' * 40)
        handshake = [
            b'hello\n',
            b'between\n',
            b'pairs %d\n' % len(pairsarg),
            pairsarg,
        ]

        if requestlog:
            ui.debug(b'devel-peer-request: hello+between\n')
            ui.debug(b'devel-peer-request:   pairs: %d bytes\n' % len(pairsarg))
        ui.debug(b'sending hello command\n')
        ui.debug(b'sending between command\n')

        stdin.write(b''.join(handshake))
        stdin.flush()
    except IOError:
        badresponse()

    # Assume version 1 of wire protocol by default.
    protoname = wireprototypes.SSHV1
    reupgraded = re.compile(b'^upgraded %s (.*)$' % stringutil.reescape(token))

    lines = [b'', b'dummy']
    max_noise = 500
    while lines[-1] and max_noise:
        try:
            l = stdout.readline()
            _forwardoutput(ui, stderr, warn=True)

            # Look for reply to protocol upgrade request. It has a token
            # in it, so there should be no false positives.
            m = reupgraded.match(l)
            if m:
                protoname = m.group(1)
                ui.debug(b'protocol upgraded to %s\n' % protoname)
                # If an upgrade was handled, the ``hello`` and ``between``
                # requests are ignored. The next output belongs to the
                # protocol, so stop scanning lines.
                break

            # Otherwise it could be a banner, ``0\n`` response if server
            # doesn't support upgrade.

            if lines[-1] == b'1\n' and l == b'\n':
                break
            if l:
                ui.debug(b'remote: ', l)
            lines.append(l)
            max_noise -= 1
        except IOError:
            badresponse()
    else:
        badresponse()

    caps = set()

    # For version 1, we should see a ``capabilities`` line in response to the
    # ``hello`` command.
    if protoname == wireprototypes.SSHV1:
        for l in reversed(lines):
            # Look for response to ``hello`` command. Scan from the back so
            # we don't misinterpret banner output as the command reply.
            if l.startswith(b'capabilities:'):
                caps.update(l[:-1].split(b':')[1].split())
                break

    # Error if we couldn't find capabilities, this means:
    #
    # 1. Remote isn't a Mercurial server
    # 2. Remote is a <0.9.1 Mercurial server
    # 3. Remote is a future Mercurial server that dropped ``hello``
    #    and other attempted handshake mechanisms.
    if not caps:
        badresponse()

    # Flush any output on stderr before proceeding.
    _forwardoutput(ui, stderr, warn=True)

    return protoname, caps


class sshv1peer(wireprotov1peer.wirepeer):
    def __init__(
        self, ui, url, proc, stdin, stdout, stderr, caps, autoreadstderr=True
    ):
        """Create a peer from an existing SSH connection.

        ``proc`` is a handle on the underlying SSH process.
        ``stdin``, ``stdout``, and ``stderr`` are handles on the stdio
        pipes for that process.
        ``caps`` is a set of capabilities supported by the remote.
        ``autoreadstderr`` denotes whether to automatically read from
        stderr and to forward its output.
        """
        self._url = url
        self.ui = ui
        # self._subprocess is unused. Keeping a handle on the process
        # holds a reference and prevents it from being garbage collected.
        self._subprocess = proc

        # And we hook up our "doublepipe" wrapper to allow querying
        # stderr any time we perform I/O.
        if autoreadstderr:
            stdout = doublepipe(ui, util.bufferedinputpipe(stdout), stderr)
            stdin = doublepipe(ui, stdin, stderr)

        self._pipeo = stdin
        self._pipei = stdout
        self._pipee = stderr
        self._caps = caps
        self._autoreadstderr = autoreadstderr
        self._initstack = b''.join(util.getstackframes(1))

    # Commands that have a "framed" response where the first line of the
    # response contains the length of that response.
    _FRAMED_COMMANDS = {
        b'batch',
    }

    # Begin of ipeerconnection interface.

    def url(self):
        return self._url

    def local(self):
        return None

    def peer(self):
        return self

    def canpush(self):
        return True

    def close(self):
        self._cleanup()

    # End of ipeerconnection interface.

    # Begin of ipeercommands interface.

    def capabilities(self):
        return self._caps

    # End of ipeercommands interface.

    def _readerr(self):
        _forwardoutput(self.ui, self._pipee)

    def _abort(self, exception):
        self._cleanup()
        raise exception

    def _cleanup(self, warn=None):
        _cleanuppipes(self.ui, self._pipei, self._pipeo, self._pipee, warn=warn)

    def __del__(self):
        self._cleanup(warn=self._initstack)

    def _sendrequest(self, cmd, args, framed=False):
        if self.ui.debugflag and self.ui.configbool(
            b'devel', b'debug.peer-request'
        ):
            dbg = self.ui.debug
            line = b'devel-peer-request: %s\n'
            dbg(line % cmd)
            for key, value in sorted(args.items()):
                if not isinstance(value, dict):
                    dbg(line % b'  %s: %d bytes' % (key, len(value)))
                else:
                    for dk, dv in sorted(value.items()):
                        dbg(line % b'  %s-%s: %d' % (key, dk, len(dv)))
        self.ui.debug(b"sending %s command\n" % cmd)
        self._pipeo.write(b"%s\n" % cmd)
        _func, names = wireprotov1server.commands[cmd]
        keys = names.split()
        wireargs = {}
        for k in keys:
            if k == b'*':
                wireargs[b'*'] = args
                break
            else:
                wireargs[k] = args[k]
                del args[k]
        for k, v in sorted(pycompat.iteritems(wireargs)):
            self._pipeo.write(b"%s %d\n" % (k, len(v)))
            if isinstance(v, dict):
                for dk, dv in pycompat.iteritems(v):
                    self._pipeo.write(b"%s %d\n" % (dk, len(dv)))
                    self._pipeo.write(dv)
            else:
                self._pipeo.write(v)
        self._pipeo.flush()

        # We know exactly how many bytes are in the response. So return a proxy
        # around the raw output stream that allows reading exactly this many
        # bytes. Callers then can read() without fear of overrunning the
        # response.
        if framed:
            amount = self._getamount()
            return util.cappedreader(self._pipei, amount)

        return self._pipei

    def _callstream(self, cmd, **args):
        args = pycompat.byteskwargs(args)
        return self._sendrequest(cmd, args, framed=cmd in self._FRAMED_COMMANDS)

    def _callcompressable(self, cmd, **args):
        args = pycompat.byteskwargs(args)
        return self._sendrequest(cmd, args, framed=cmd in self._FRAMED_COMMANDS)

    def _call(self, cmd, **args):
        args = pycompat.byteskwargs(args)
        return self._sendrequest(cmd, args, framed=True).read()

    def _callpush(self, cmd, fp, **args):
        # The server responds with an empty frame if the client should
        # continue submitting the payload.
        r = self._call(cmd, **args)
        if r:
            return b'', r

        # The payload consists of frames with content followed by an empty
        # frame.
        for d in iter(lambda: fp.read(4096), b''):
            self._writeframed(d)
        self._writeframed(b"", flush=True)

        # In case of success, there is an empty frame and a frame containing
        # the integer result (as a string).
        # In case of error, there is a non-empty frame containing the error.
        r = self._readframed()
        if r:
            return b'', r
        return self._readframed(), b''

    def _calltwowaystream(self, cmd, fp, **args):
        # The server responds with an empty frame if the client should
        # continue submitting the payload.
        r = self._call(cmd, **args)
        if r:
            # XXX needs to be made better
            raise error.Abort(_(b'unexpected remote reply: %s') % r)

        # The payload consists of frames with content followed by an empty
        # frame.
        for d in iter(lambda: fp.read(4096), b''):
            self._writeframed(d)
        self._writeframed(b"", flush=True)

        return self._pipei

    def _getamount(self):
        l = self._pipei.readline()
        if l == b'\n':
            if self._autoreadstderr:
                self._readerr()
            msg = _(b'check previous remote output')
            self._abort(error.OutOfBandError(hint=msg))
        if self._autoreadstderr:
            self._readerr()
        try:
            return int(l)
        except ValueError:
            self._abort(error.ResponseError(_(b"unexpected response:"), l))

    def _readframed(self):
        size = self._getamount()
        if not size:
            return b''

        return self._pipei.read(size)

    def _writeframed(self, data, flush=False):
        self._pipeo.write(b"%d\n" % len(data))
        if data:
            self._pipeo.write(data)
        if flush:
            self._pipeo.flush()
        if self._autoreadstderr:
            self._readerr()


def makepeer(ui, path, proc, stdin, stdout, stderr, autoreadstderr=True):
    """Make a peer instance from existing pipes.

    ``path`` and ``proc`` are stored on the eventual peer instance and may
    not be used for anything meaningful.

    ``stdin``, ``stdout``, and ``stderr`` are the pipes connected to the
    SSH server's stdio handles.

    This function is factored out to allow creating peers that don't
    actually spawn a new process. It is useful for starting SSH protocol
    servers and clients via non-standard means, which can be useful for
    testing.
    """
    try:
        protoname, caps = _performhandshake(ui, stdin, stdout, stderr)
    except Exception:
        _cleanuppipes(ui, stdout, stdin, stderr, warn=None)
        raise

    if protoname == wireprototypes.SSHV1:
        return sshv1peer(
            ui,
            path,
            proc,
            stdin,
            stdout,
            stderr,
            caps,
            autoreadstderr=autoreadstderr,
        )
    else:
        _cleanuppipes(ui, stdout, stdin, stderr, warn=None)
        raise error.RepoError(
            _(b'unknown version of SSH protocol: %s') % protoname
        )


def instance(ui, path, create, intents=None, createopts=None):
    """Create an SSH peer.

    The returned object conforms to the ``wireprotov1peer.wirepeer`` interface.
    """
    u = urlutil.url(path, parsequery=False, parsefragment=False)
    if u.scheme != b'ssh' or not u.host or u.path is None:
        raise error.RepoError(_(b"couldn't parse location %s") % path)

    urlutil.checksafessh(path)

    if u.passwd is not None:
        raise error.RepoError(_(b'password in URL not supported'))

    sshcmd = ui.config(b'ui', b'ssh')
    remotecmd = ui.config(b'ui', b'remotecmd')
    sshaddenv = dict(ui.configitems(b'sshenv'))
    sshenv = procutil.shellenviron(sshaddenv)
    remotepath = u.path or b'.'

    args = procutil.sshargs(sshcmd, u.host, u.user, u.port)

    if create:
        # We /could/ do this, but only if the remote init command knows how to
        # handle them. We don't yet make any assumptions about that. And without
        # querying the remote, there's no way of knowing if the remote even
        # supports said requested feature.
        if createopts:
            raise error.RepoError(
                _(
                    b'cannot create remote SSH repositories '
                    b'with extra options'
                )
            )

        cmd = b'%s %s %s' % (
            sshcmd,
            args,
            procutil.shellquote(
                b'%s init %s'
                % (_serverquote(remotecmd), _serverquote(remotepath))
            ),
        )
        ui.debug(b'running %s\n' % cmd)
        res = ui.system(cmd, blockedtag=b'sshpeer', environ=sshenv)
        if res != 0:
            raise error.RepoError(_(b'could not create remote repo'))

    proc, stdin, stdout, stderr = _makeconnection(
        ui, sshcmd, args, remotecmd, remotepath, sshenv
    )

    peer = makepeer(ui, path, proc, stdin, stdout, stderr)

    # Finally, if supported by the server, notify it about our own
    # capabilities.
    if b'protocaps' in peer.capabilities():
        try:
            peer._call(
                b"protocaps", caps=b' '.join(sorted(_clientcapabilities()))
            )
        except IOError:
            peer._cleanup()
            raise error.RepoError(_(b'capability exchange failed'))

    return peer