'''

Module provides a class allowing to wrap communication over subprocess.Popen

input, output, error streams into a meaningfull, non-blocking, concurrent

stream processor exposing the output data as an iterator fitting to be a

return value passed by a WSGI applicaiton to a WSGI server per PEP 3333.

Copyright (c) 2011  Daniel Dotsenko <dotsa@hotmail.com>

This file is part of git_http_backend.py Project.

git_http_backend.py Project is free software: you can redistribute it and/or

modify it under the terms of the GNU Lesser General Public License as

published by the Free Software Foundation, either version 2.1 of the License,

or (at your option) any later version.

git_http_backend.py Project 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 Lesser General Public License for more details.

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

along with git_http_backend.py Project.

If not, see <http://www.gnu.org/licenses/>.

'''

import os

import subprocess

from rhodecode.lib.compat import deque, Event, Thread, _bytes, _bytearray

class StreamFeeder(Thread):

    """

    Normal writing into pipe-like is blocking once the buffer is filled.

    This thread allows a thread to seep data from a file-like into a pipe

    without blocking the main thread.

    We close inpipe once the end of the source stream is reached.

    """

    def __init__(self, source):

        super(StreamFeeder, self).__init__()

        self.daemon = True

        filelike = False

        self.bytes = _bytes()

        if type(source) in (type(''), _bytes, _bytearray):  # string-like

            self.bytes = _bytes(source)

        else:  # can be either file pointer or file-like

            if type(source) in (int, long):  # file pointer it is

                ## converting file descriptor (int) stdin into file-like

                try:

                    source = os.fdopen(source, 'rb', 16384)

                except Exception:

                    pass

            # let's see if source is file-like by now

            try:

                filelike = source.read

            except Exception:

                pass

        if not filelike and not self.bytes:

            raise TypeError("StreamFeeder's source object must be a readable "

                            "file-like, a file descriptor, or a string-like.")

        self.source = source

        self.readiface, self.writeiface = os.pipe()

    def run(self):

        t = self.writeiface

        if self.bytes:

            os.write(t, self.bytes)

        else:

            s = self.source

            b = s.read(4096)

            while b:

                os.write(t, b)

                b = s.read(4096)

        os.close(t)

    @property

    def output(self):

        return self.readiface

class InputStreamChunker(Thread):

    def __init__(self, source, target, buffer_size, chunk_size):

        super(InputStreamChunker, self).__init__()

        self.daemon = True  # die die die.

        self.source = source

        self.target = target

        self.chunk_count_max = int(buffer_size / chunk_size) + 1

        self.chunk_size = chunk_size

        self.data_added = Event()

        self.data_added.clear()

        self.keep_reading = Event()

        self.keep_reading.set()

        self.EOF = Event()

        self.EOF.clear()

        self.go = Event()

        self.go.set()

    def stop(self):

        self.go.clear()

        self.EOF.set()

        try:

            # this is not proper, but is done to force the reader thread let

            # go of the input because, if successful, .close() will send EOF

            # down the pipe.

            self.source.close()

        except:

            pass

    def run(self):

        s = self.source

        t = self.target

        cs = self.chunk_size

        ccm = self.chunk_count_max

        kr = self.keep_reading

        da = self.data_added

        go = self.go

        b = s.read(cs)

        while b and go.is_set():

            if len(t) > ccm:

                kr.clear()

                kr.wait(2)

#                # this only works on 2.7.x and up

#                if not kr.wait(10):

#                    raise Exception("Timed out while waiting for input to be read.")

                # instead we'll use this

                if len(t) > ccm + 3:

                    raise IOError("Timed out while waiting for input from subprocess.")

            t.append(b)

            da.set()

            b = s.read(cs)

        self.EOF.set()

        da.set()  # for cases when done but there was no input.

class BufferedGenerator():

    '''

    Class behaves as a non-blocking, buffered pipe reader.

    Reads chunks of data (through a thread)

    from a blocking pipe, and attaches these to an array (Deque) of chunks.

    Reading is halted in the thread when max chunks is internally buffered.

    The .next() may operate in blocking or non-blocking fashion by yielding

    '' if no data is ready

    to be sent or by not returning until there is some data to send

    When we get EOF from underlying source pipe we raise the marker to raise

    StopIteration after the last chunk of data is yielded.

    '''

    def __init__(self, source, buffer_size=65536, chunk_size=4096,

                 starting_values=[], bottomless=False):

        if bottomless:

            maxlen = int(buffer_size / chunk_size)

        else:

            maxlen = None

        self.data = deque(starting_values, maxlen)

        self.worker = InputStreamChunker(source, self.data, buffer_size,

                                         chunk_size)

        if starting_values:

            self.worker.data_added.set()

        self.worker.start()

    ####################

    # Generator's methods

    ####################

    def __iter__(self):

        return self

    def next(self):

        while not len(self.data) and not self.worker.EOF.is_set():

            self.worker.data_added.clear()

            self.worker.data_added.wait(0.2)

        if len(self.data):

            self.worker.keep_reading.set()

            return _bytes(self.data.popleft())

        elif self.worker.EOF.is_set():

            raise StopIteration

    def throw(self, type, value=None, traceback=None):

        if not self.worker.EOF.is_set():

            raise type(value)

    def start(self):

        self.worker.start()

    def stop(self):

        self.worker.stop()

    def close(self):

        try:

            self.worker.stop()

            self.throw(GeneratorExit)

        except (GeneratorExit, StopIteration):

            pass

    def __del__(self):

        self.close()

    ####################

    # Threaded reader's infrastructure.

    ####################

    @property

    def input(self):

        return self.worker.w

    @property

    def data_added_event(self):

        return self.worker.data_added

    @property

    def data_added(self):

        return self.worker.data_added.is_set()

    @property

    def reading_paused(self):

        return not self.worker.keep_reading.is_set()

    @property

    def done_reading_event(self):

        '''

        Done_reding does not mean that the iterator's buffer is empty.

        Iterator might have done reading from underlying source, but the read

        chunks might still be available for serving through .next() method.

        @return An Event class instance.

        '''

        return self.worker.EOF

    @property

    def done_reading(self):

        '''

        Done_reding does not mean that the iterator's buffer is empty.

        Iterator might have done reading from underlying source, but the read

        chunks might still be available for serving through .next() method.

        @return An Bool value.

        '''

        return self.worker.EOF.is_set()

    @property

    def length(self):

        '''

        returns int.

        This is the lenght of the que of chunks, not the length of

        the combined contents in those chunks.

        __len__() cannot be meaningfully implemented because this

        reader is just flying throuh a bottomless pit content and

        can only know the lenght of what it already saw.

        If __len__() on WSGI server per PEP 3333 returns a value,

        the responce's length will be set to that. In order not to

        confuse WSGI PEP3333 servers, we will not implement __len__

        at all.

        '''

        return len(self.data)

    def prepend(self, x):

        self.data.appendleft(x)

    def append(self, x):

        self.data.append(x)

    def extend(self, o):

        self.data.extend(o)

    def __getitem__(self, i):

        return self.data[i]

class SubprocessIOChunker(object):

    '''

    Processor class wrapping handling of subprocess IO.

    In a way, this is a "communicate()" replacement with a twist.

    - We are multithreaded. Writing in and reading out, err are all sep threads.

    - We support concurrent (in and out) stream processing.

    - The output is not a stream. It's a queue of read string (bytes, not unicode)

      chunks. The object behaves as an iterable. You can "for chunk in obj:" us.

    - We are non-blocking in more respects than communicate()

      (reading from subprocess out pauses when internal buffer is full, but

       does not block the parent calling code. On the flip side, reading from

       slow-yielding subprocess may block the iteration until data shows up. This

       does not block the parallel inpipe reading occurring parallel thread.)

    The purpose of the object is to allow us to wrap subprocess interactions into

    and interable that can be passed to a WSGI server as the application's return

    value. Because of stream-processing-ability, WSGI does not have to read ALL

    of the subprocess's output and buffer it, before handing it to WSGI server for

    HTTP response. Instead, the class initializer reads just a bit of the stream

    to figure out if error ocurred or likely to occur and if not, just hands the

    further iteration over subprocess output to the server for completion of HTTP

    response.

    The real or perceived subprocess error is trapped and raised as one of

    EnvironmentError family of exceptions

    Example usage:

    #    try:

    #        answer = SubprocessIOChunker(

    #            cmd,

    #            input,

    #            buffer_size = 65536,

    #            chunk_size = 4096

    #            )

    #    except (EnvironmentError) as e:

    #        print str(e)

    #        raise e

    #

    #    return answer

    '''

    def __init__(self, cmd, inputstream=None, buffer_size=65536,

                 chunk_size=4096, starting_values=[], **kwargs):

        '''

        Initializes SubprocessIOChunker

        :param cmd: A Subprocess.Popen style "cmd". Can be string or array of strings

        :param inputstream: (Default: None) A file-like, string, or file pointer.

        :param buffer_size: (Default: 65536) A size of total buffer per stream in bytes.

        :param chunk_size: (Default: 4096) A max size of a chunk. Actual chunk may be smaller.

        :param starting_values: (Default: []) An array of strings to put in front of output que.

        '''

        if inputstream:

            input_streamer = StreamFeeder(inputstream)

            input_streamer.start()

            inputstream = input_streamer.output

        if isinstance(cmd, (list, tuple)):

            cmd = ' '.join(cmd)

        _shell = kwargs.get('shell') or True

        kwargs['shell'] = _shell

        _p = subprocess.Popen(cmd,

            bufsize=-1,

            stdin=inputstream,

            stdout=subprocess.PIPE,

            stderr=subprocess.PIPE,

            **kwargs

            )

        bg_out = BufferedGenerator(_p.stdout, buffer_size, chunk_size, starting_values)

        bg_err = BufferedGenerator(_p.stderr, 16000, 1, bottomless=True)

        while not bg_out.done_reading and not bg_out.reading_paused and not bg_err.length:

            # doing this until we reach either end of file, or end of buffer.

            bg_out.data_added_event.wait(1)

            bg_out.data_added_event.clear()

        # at this point it's still ambiguous if we are done reading or just full buffer.

        # Either way, if error (returned by ended process, or implied based on

        # presence of stuff in stderr output) we error out.

        # Else, we are happy.

        _returncode = _p.poll()

        if _returncode or (_returncode == None and bg_err.length):

            try:

                _p.terminate()

            except:

                pass

            bg_out.stop()

            bg_err.stop()

            err = '%s' % ''.join(bg_err)

            raise EnvironmentError("Subprocess exited due to an error:\n" + err)

        self.process = _p

        self.output = bg_out

        self.error = bg_err

    def __iter__(self):

        return self

    def next(self):

        if self.process.poll():

            err = '%s' % ''.join(self.error)

            raise EnvironmentError("Subprocess exited due to an error:\n" + err)

        return self.output.next()

    def throw(self, type, value=None, traceback=None):

        if self.output.length or not self.output.done_reading:

            raise type(value)

    def close(self):

        try:

            self.process.terminate()

        except:

            pass

        try:

            self.output.close()

        except:

            pass

        try:

            self.error.close()

        except:

            pass

    def __del__(self):

        self.close()

def get_custom_lexer(extension):

    """

    returns a custom lexer if it's defined in rcextensions module, or None

    if there's no custom lexer defined

    """

    import rhodecode

    from pygments import lexers

    #check if we didn't define this extension as other lexer

    if rhodecode.EXTENSIONS and extension in rhodecode.EXTENSIONS.EXTRA_LEXERS:

        _lexer_name = rhodecode.EXTENSIONS.EXTRA_LEXERS[extension]

        return lexers.get_lexer_by_name(_lexer_name)

#==============================================================================

# TEST FUNCTIONS AND CREATORS

#==============================================================================

def create_test_index(repo_location, config, full_index):

    """

    Makes default test index

    :param config: test config

    :param full_index:

    """

    from rhodecode.lib.indexers.daemon import WhooshIndexingDaemon

    from rhodecode.lib.pidlock import DaemonLock, LockHeld

    repo_location = repo_location

    index_location = os.path.join(config['app_conf']['index_dir'])

    if not os.path.exists(index_location):

        os.makedirs(index_location)

    try:

        l = DaemonLock(file_=jn(dn(index_location), 'make_index.lock'))

        WhooshIndexingDaemon(index_location=index_location,

                             repo_location=repo_location)\

            .run(full_index=full_index)

        l.release()

    except LockHeld:

        pass

def create_test_env(repos_test_path, config):

    """

    Makes a fresh database and

    install test repository into tmp dir

    """

    from rhodecode.lib.db_manage import DbManage

    from rhodecode.tests import HG_REPO, GIT_REPO, TESTS_TMP_PATH

    # PART ONE create db

    dbconf = config['sqlalchemy.db1.url']

    log.debug('making test db %s' % dbconf)

    # create test dir if it doesn't exist

    if not os.path.isdir(repos_test_path):

        log.debug('Creating testdir %s' % repos_test_path)

        os.makedirs(repos_test_path)

    dbmanage = DbManage(log_sql=True, dbconf=dbconf, root=config['here'],

                        tests=True)

    dbmanage.create_tables(override=True)

    dbmanage.create_settings(dbmanage.config_prompt(repos_test_path))

    dbmanage.create_default_user()

    dbmanage.admin_prompt()

    dbmanage.create_permissions()

    dbmanage.populate_default_permissions()

    Session().commit()

    # PART TWO make test repo

    log.debug('making test vcs repositories')

    idx_path = config['app_conf']['index_dir']

    data_path = config['app_conf']['cache_dir']

    #clean index and data

    if idx_path and os.path.exists(idx_path):

        log.debug('remove %s' % idx_path)

        shutil.rmtree(idx_path)

    if data_path and os.path.exists(data_path):

        log.debug('remove %s' % data_path)

        shutil.rmtree(data_path)

    #CREATE DEFAULT TEST REPOS

    cur_dir = dn(dn(abspath(__file__)))

    tar = tarfile.open(jn(cur_dir, 'tests', "vcs_test_hg.tar.gz"))

    tar.extractall(jn(TESTS_TMP_PATH, HG_REPO))

    tar.close()

    cur_dir = dn(dn(abspath(__file__)))

    tar = tarfile.open(jn(cur_dir, 'tests', "vcs_test_git.tar.gz"))

    tar.extractall(jn(TESTS_TMP_PATH, GIT_REPO))

    tar.close()

    #LOAD VCS test stuff

    from rhodecode.tests.vcs import setup_package

    setup_package()

#==============================================================================

# PASTER COMMANDS

#==============================================================================

class BasePasterCommand(Command):

    """

    Abstract Base Class for paster commands.

    The celery commands are somewhat aggressive about loading

    celery.conf, and since our module sets the `CELERY_LOADER`

    environment variable to our loader, we have to bootstrap a bit and

    make sure we've had a chance to load the pylons config off of the

    command line, otherwise everything fails.

    """

    min_args = 1

    min_args_error = "Please provide a paster config file as an argument."

    takes_config_file = 1

    requires_config_file = True

    def notify_msg(self, msg, log=False):

        """Make a notification to user, additionally if logger is passed

        it logs this action using given logger

        :param msg: message that will be printed to user

        :param log: logging instance, to use to additionally log this message

        """

        if log and isinstance(log, logging):

            log(msg)

    def run(self, args):

        """

        Overrides Command.run

        Checks for a config file argument and loads it.

        """

        if len(args) < self.min_args:

            raise BadCommand(

                self.min_args_error % {'min_args': self.min_args,

                                       'actual_args': len(args)})

        # Decrement because we're going to lob off the first argument.

        # @@ This is hacky

        self.min_args -= 1

        self.bootstrap_config(args[0])

        self.update_parser()

        return super(BasePasterCommand, self).run(args[1:])

    def update_parser(self):

        """

        Abstract method.  Allows for the class's parser to be updated

        before the superclass's `run` method is called.  Necessary to

        allow options/arguments to be passed through to the underlying

        celery command.

        """

        raise NotImplementedError("Abstract Method.")

    def bootstrap_config(self, conf):

        """

        Loads the pylons configuration.

        """

        from pylons import config as pylonsconfig

        self.path_to_ini_file = os.path.realpath(conf)

        conf = paste.deploy.appconfig('config:' + self.path_to_ini_file)

        pylonsconfig.init_app(conf.global_conf, conf.local_conf)

    def _init_session(self):

        """

        Inits SqlAlchemy Session

        """

        logging.config.fileConfig(self.path_to_ini_file)

        from pylons import config

        from rhodecode.model import init_model

        from rhodecode.lib.utils2 import engine_from_config

        #get to remove repos !!

        add_cache(config)

        engine = engine_from_config(config, 'sqlalchemy.db1.')

        init_model(engine)

def check_git_version():

    """

    Checks what version of git is installed in system, and issues a warning

    if it's too old for RhodeCode to properly work.

    """

    from rhodecode import BACKENDS

    from rhodecode.lib.vcs.backends.git.repository import GitRepository

    from distutils.version import StrictVersion

    ver = (stdout.split(' ')[-1] or '').strip() or '0.0.0'

    if len(ver.split('.')) > 3:

        #StrictVersion needs to be only 3 element type

        ver = '.'.join(ver.split('.')[:3])

    try:

        _ver = StrictVersion(ver)

    except:

        _ver = StrictVersion('0.0.0')

        stderr = traceback.format_exc()

    req_ver = '1.7.4'

    to_old_git = False

    if  _ver < StrictVersion(req_ver):

        to_old_git = True

    if 'git' in BACKENDS:

        log.debug('GIT version detected: %s' % stdout)

        if stderr:

            log.warning('Unable to detect git version org error was:%r' % stderr)

        elif to_old_git:

            log.warning('RhodeCode detected git version %s, which is too old '

                        'for the system to function properly. Make sure '

                        'its version is at least %s' % (ver, req_ver))

    return _ver

@decorator.decorator

def jsonify(func, *args, **kwargs):

    """Action decorator that formats output for JSON

    Given a function that will return content, this decorator will turn

    the result into JSON, with a content-type of 'application/json' and

    output it.

    """

    from pylons.decorators.util import get_pylons

    from rhodecode.lib.ext_json import json

    pylons = get_pylons(args)

    pylons.response.headers['Content-Type'] = 'application/json; charset=utf-8'

    data = func(*args, **kwargs)

    if isinstance(data, (list, tuple)):

        msg = "JSON responses with Array envelopes are susceptible to " \

              "cross-site data leak attacks, see " \

              "http://wiki.pylonshq.com/display/pylonsfaq/Warnings"

        warnings.warn(msg, Warning, 2)

        log.warning(msg)

    log.debug("Returning JSON wrapped action output")

    return json.dumps(data, encoding='utf-8')

# -*- coding: utf-8 -*-

"""

    vcs.backends.git

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

    Git backend implementation.

    :created_on: Apr 8, 2010

    :copyright: (c) 2010-2011 by Marcin Kuzminski, Lukasz Balcerzak.

"""

import os

import re

import time

import posixpath

import logging

import traceback

import urllib

import urllib2

from dulwich.repo import Repo, NotGitRepository

from dulwich.objects import Tag

from string import Template

import rhodecode

from rhodecode.lib.vcs.backends.base import BaseRepository

from rhodecode.lib.vcs.exceptions import BranchDoesNotExistError

from rhodecode.lib.vcs.exceptions import ChangesetDoesNotExistError

from rhodecode.lib.vcs.exceptions import EmptyRepositoryError

from rhodecode.lib.vcs.exceptions import RepositoryError

from rhodecode.lib.vcs.exceptions import TagAlreadyExistError

from rhodecode.lib.vcs.exceptions import TagDoesNotExistError

from rhodecode.lib.vcs.utils import safe_unicode, makedate, date_fromtimestamp

from rhodecode.lib.vcs.utils.lazy import LazyProperty, ThreadLocalLazyProperty

from rhodecode.lib.vcs.utils.ordered_dict import OrderedDict

from rhodecode.lib.vcs.utils.paths import abspath

from rhodecode.lib.vcs.utils.paths import get_user_home

from .workdir import GitWorkdir

from .changeset import GitChangeset

from .inmemory import GitInMemoryChangeset

from .config import ConfigFile

from rhodecode.lib import subprocessio

log = logging.getLogger(__name__)

class GitRepository(BaseRepository):

    """

    Git repository backend.

    """

    DEFAULT_BRANCH_NAME = 'master'

    scm = 'git'

    def __init__(self, repo_path, create=False, src_url=None,

                 update_after_clone=False, bare=False):

        self.path = abspath(repo_path)

        repo = self._get_repo(create, src_url, update_after_clone, bare)

        self.bare = repo.bare

        self._config_files = [

            bare and abspath(self.path, 'config')

                     or abspath(self.path, '.git', 'config'),

            abspath(get_user_home(), '.gitconfig'),

        ]

    @ThreadLocalLazyProperty

    def _repo(self):

        repo = Repo(self.path)

        #temporary set that to now at later we will move it to constructor

        baseui = None

        if baseui is None:

            from mercurial.ui import ui

            baseui = ui()

        # patch the instance of GitRepo with an "FAKE" ui object to add

        # compatibility layer with Mercurial

        setattr(repo, 'ui', baseui)

        return repo

    @property

    def head(self):

        try:

            return self._repo.head()

        except KeyError:

            return None

    @LazyProperty

    def revisions(self):

        """

        Returns list of revisions' ids, in ascending order.  Being lazy

        attribute allows external tools to inject shas from cache.

        """

        return self._get_all_revisions()

    @classmethod

    def _run_git_command(cls, cmd, **opts):

        """

        Runs given ``cmd`` as git command and returns tuple

        (stdout, stderr).

        :param cmd: git command to be executed

        :param opts: env options to pass into Subprocess command

        """

        _copts = ['-c', 'core.quotepath=false', ]

        _str_cmd = False

        if isinstance(cmd, basestring):

            cmd = [cmd]

            _str_cmd = True

        gitenv = os.environ

        # need to clean fix GIT_DIR !

        if 'GIT_DIR' in gitenv:

            del gitenv['GIT_DIR']

        gitenv['GIT_CONFIG_NOGLOBAL'] = '1'

        _git_path = rhodecode.CONFIG.get('git_path', 'git')

        cmd = [_git_path] + _copts + cmd

        if _str_cmd:

            cmd = ' '.join(cmd)

        try:

            _opts = dict(

                env=gitenv,

                shell=False,

            )

            _opts.update(opts)

            p = subprocessio.SubprocessIOChunker(cmd, **_opts)

        except (EnvironmentError, OSError), err:

        return ''.join(p.output), ''.join(p.error)

    def run_git_command(self, cmd):

        opts = {}

        if os.path.isdir(self.path):

            opts['cwd'] = self.path

        return self._run_git_command(cmd, **opts)

    @classmethod

    def _check_url(cls, url):

        """

        Functon will check given url and try to verify if it's a valid

        link. Sometimes it may happened that mercurial will issue basic

        auth request that can cause whole API to hang when used from python

        or other external calls.

        On failures it'll raise urllib2.HTTPError

        """

        from mercurial.util import url as Url

        # those authnadlers are patched for python 2.6.5 bug an

        # infinit looping when given invalid resources

        from mercurial.url import httpbasicauthhandler, httpdigestauthhandler

        # check first if it's not an local url

        if os.path.isdir(url) or url.startswith('file:'):

            return True

        if('+' in url[:url.find('://')]):

            url = url[url.find('+') + 1:]

        handlers = []

        test_uri, authinfo = Url(url).authinfo()

        if not test_uri.endswith('info/refs'):

            test_uri = test_uri.rstrip('/') + '/info/refs'

        if authinfo:

            #create a password manager

            passmgr = urllib2.HTTPPasswordMgrWithDefaultRealm()

            passmgr.add_password(*authinfo)

            handlers.extend((httpbasicauthhandler(passmgr),

                             httpdigestauthhandler(passmgr)))

        o = urllib2.build_opener(*handlers)

        o.addheaders = [('User-Agent', 'git/1.7.8.0')]  # fake some git

        q = {"service": 'git-upload-pack'}

        qs = '?%s' % urllib.urlencode(q)

        cu = "%s%s" % (test_uri, qs)

        req = urllib2.Request(cu, None, {})

        try:

            resp = o.open(req)

            return resp.code == 200

        except Exception, e:

            # means it cannot be cloned

            raise urllib2.URLError("[%s] %s" % (url, e))

    def _get_repo(self, create, src_url=None, update_after_clone=False,

                  bare=False):

        if create and os.path.exists(self.path):