linting

vcsserver/hooks.py

@@ -1,738 +1,738 b''
 # -*- coding: utf-8 -*-
 
 # RhodeCode VCSServer provides access to different vcs backends via network.
 # Copyright (C) 2014-2020 RhodeCode GmbH
 #
 # 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 3 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301  USA
 
 import io
 import os
 import sys
 import logging
 import collections
 import importlib
 import base64
 import msgpack
 
 from http.client import HTTPConnection
 
 
 import mercurial.scmutil
 import mercurial.node
 
 from vcsserver.lib.rc_json import json
 from vcsserver import exceptions, subprocessio, settings
 from vcsserver.str_utils import safe_bytes
 
 log = logging.getLogger(__name__)
 
 
 class HooksHttpClient(object):
     proto = 'msgpack.v1'
     connection = None
 
     def __init__(self, hooks_uri):
         self.hooks_uri = hooks_uri
 
     def __call__(self, method, extras):
         connection = HTTPConnection(self.hooks_uri)
         # binary msgpack body
         headers, body = self._serialize(method, extras)
         try:
             connection.request('POST', '/', body, headers)
         except Exception as error:
             log.error('Hooks calling Connection failed on %s, org error: %s', connection.__dict__, error)
             raise
         response = connection.getresponse()
         try:
             return msgpack.load(response)
         except Exception:
             response_data = response.read()
             log.exception('Failed to decode hook response json data. '
                           'response_code:%s, raw_data:%s',
                           response.status, response_data)
             raise
 
     @classmethod
     def _serialize(cls, hook_name, extras):
         data = {
             'method': hook_name,
             'extras': extras
         }
         headers = {
             'rc-hooks-protocol': cls.proto
         }
         return headers, msgpack.packb(data)
 
 
 class HooksDummyClient(object):
     def __init__(self, hooks_module):
         self._hooks_module = importlib.import_module(hooks_module)
 
     def __call__(self, hook_name, extras):
         with self._hooks_module.Hooks() as hooks:
             return getattr(hooks, hook_name)(extras)
 
 
 class HooksShadowRepoClient(object):
 
     def __call__(self, hook_name, extras):
         return {'output': '', 'status': 0}
 
 
 class RemoteMessageWriter(object):
     """Writer base class."""
     def write(self, message):
         raise NotImplementedError()
 
 
 class HgMessageWriter(RemoteMessageWriter):
     """Writer that knows how to send messages to mercurial clients."""
 
     def __init__(self, ui):
         self.ui = ui
 
     def write(self, message):
         # TODO: Check why the quiet flag is set by default.
         old = self.ui.quiet
         self.ui.quiet = False
         self.ui.status(message.encode('utf-8'))
         self.ui.quiet = old
 
 
 class GitMessageWriter(RemoteMessageWriter):
     """Writer that knows how to send messages to git clients."""
 
     def __init__(self, stdout=None):
         self.stdout = stdout or sys.stdout
 
     def write(self, message):
         self.stdout.write(safe_bytes(message))
 
 
 class SvnMessageWriter(RemoteMessageWriter):
     """Writer that knows how to send messages to svn clients."""
 
     def __init__(self, stderr=None):
         # SVN needs data sent to stderr for back-to-client messaging
         self.stderr = stderr or sys.stderr
 
     def write(self, message):
         self.stderr.write(message.encode('utf-8'))
 
 
 def _handle_exception(result):
     exception_class = result.get('exception')
     exception_traceback = result.get('exception_traceback')
 
     if exception_traceback:
         log.error('Got traceback from remote call:%s', exception_traceback)
 
     if exception_class == 'HTTPLockedRC':
         raise exceptions.RepositoryLockedException()(*result['exception_args'])
     elif exception_class == 'HTTPBranchProtected':
         raise exceptions.RepositoryBranchProtectedException()(*result['exception_args'])
     elif exception_class == 'RepositoryError':
         raise exceptions.VcsException()(*result['exception_args'])
     elif exception_class:
         raise Exception('Got remote exception "%s" with args "%s"' %
                         (exception_class, result['exception_args']))
 
 
 def _get_hooks_client(extras):
     hooks_uri = extras.get('hooks_uri')
     is_shadow_repo = extras.get('is_shadow_repo')
     if hooks_uri:
         return HooksHttpClient(extras['hooks_uri'])
     elif is_shadow_repo:
         return HooksShadowRepoClient()
     else:
         return HooksDummyClient(extras['hooks_module'])
 
 
 def _call_hook(hook_name, extras, writer):
     hooks_client = _get_hooks_client(extras)
     log.debug('Hooks, using client:%s', hooks_client)
     result = hooks_client(hook_name, extras)
     log.debug('Hooks got result: %s', result)
 
     _handle_exception(result)
     writer.write(result['output'])
 
     return result['status']
 
 
 def _extras_from_ui(ui):
     hook_data = ui.config(b'rhodecode', b'RC_SCM_DATA')
     if not hook_data:
         # maybe it's inside environ ?
         env_hook_data = os.environ.get('RC_SCM_DATA')
         if env_hook_data:
             hook_data = env_hook_data
 
     extras = {}
     if hook_data:
         extras = json.loads(hook_data)
     return extras
 
 
 def _rev_range_hash(repo, node, check_heads=False):
     from vcsserver.hgcompat import get_ctx
 
     commits = []
     revs = []
     start = get_ctx(repo, node).rev()
     end = len(repo)
     for rev in range(start, end):
         revs.append(rev)
         ctx = get_ctx(repo, rev)
         commit_id = mercurial.node.hex(ctx.node())
         branch = ctx.branch()
         commits.append((commit_id, branch))
 
     parent_heads = []
     if check_heads:
         parent_heads = _check_heads(repo, start, end, revs)
     return commits, parent_heads
 
 
 def _check_heads(repo, start, end, commits):
     from vcsserver.hgcompat import get_ctx
     changelog = repo.changelog
     parents = set()
 
     for new_rev in commits:
         for p in changelog.parentrevs(new_rev):
             if p == mercurial.node.nullrev:
                 continue
             if p < start:
                 parents.add(p)
 
     for p in parents:
         branch = get_ctx(repo, p).branch()
         # The heads descending from that parent, on the same branch
         parent_heads = set([p])
         reachable = set([p])
         for x in range(p + 1, end):
             if get_ctx(repo, x).branch() != branch:
                 continue
             for pp in changelog.parentrevs(x):
                 if pp in reachable:
                     reachable.add(x)
                     parent_heads.discard(pp)
                     parent_heads.add(x)
         # More than one head? Suggest merging
         if len(parent_heads) > 1:
             return list(parent_heads)
 
     return []
 
 
 def _get_git_env():
     env = {}
     for k, v in os.environ.items():
         if k.startswith('GIT'):
             env[k] = v
 
     # serialized version
     return [(k, v) for k, v in env.items()]
 
 
 def _get_hg_env(old_rev, new_rev, txnid, repo_path):
     env = {}
     for k, v in os.environ.items():
         if k.startswith('HG'):
             env[k] = v
 
     env['HG_NODE'] = old_rev
     env['HG_NODE_LAST'] = new_rev
     env['HG_TXNID'] = txnid
     env['HG_PENDING'] = repo_path
 
     return [(k, v) for k, v in env.items()]
 
 
 def repo_size(ui, repo, **kwargs):
     extras = _extras_from_ui(ui)
     return _call_hook('repo_size', extras, HgMessageWriter(ui))
 
 
 def pre_pull(ui, repo, **kwargs):
     extras = _extras_from_ui(ui)
     return _call_hook('pre_pull', extras, HgMessageWriter(ui))
 
 
 def pre_pull_ssh(ui, repo, **kwargs):
     extras = _extras_from_ui(ui)
     if extras and extras.get('SSH'):
         return pre_pull(ui, repo, **kwargs)
     return 0
 
 
 def post_pull(ui, repo, **kwargs):
     extras = _extras_from_ui(ui)
     return _call_hook('post_pull', extras, HgMessageWriter(ui))
 
 
 def post_pull_ssh(ui, repo, **kwargs):
     extras = _extras_from_ui(ui)
     if extras and extras.get('SSH'):
         return post_pull(ui, repo, **kwargs)
     return 0
 
 
 def pre_push(ui, repo, node=None, **kwargs):
     """
     Mercurial pre_push hook
     """
     extras = _extras_from_ui(ui)
     detect_force_push = extras.get('detect_force_push')
 
     rev_data = []
     if node and kwargs.get('hooktype') == 'pretxnchangegroup':
         branches = collections.defaultdict(list)
         commits, _heads = _rev_range_hash(repo, node, check_heads=detect_force_push)
         for commit_id, branch in commits:
             branches[branch].append(commit_id)
 
         for branch, commits in branches.items():
             old_rev = kwargs.get('node_last') or commits[0]
             rev_data.append({
                 'total_commits': len(commits),
                 'old_rev': old_rev,
                 'new_rev': commits[-1],
                 'ref': '',
                 'type': 'branch',
                 'name': branch,
             })
 
         for push_ref in rev_data:
             push_ref['multiple_heads'] = _heads
 
             repo_path = os.path.join(
                 extras.get('repo_store', ''), extras.get('repository', ''))
             push_ref['hg_env'] = _get_hg_env(
                 old_rev=push_ref['old_rev'],
                 new_rev=push_ref['new_rev'], txnid=kwargs.get('txnid'),
                 repo_path=repo_path)
 
     extras['hook_type'] = kwargs.get('hooktype', 'pre_push')
     extras['commit_ids'] = rev_data
 
     return _call_hook('pre_push', extras, HgMessageWriter(ui))
 
 
 def pre_push_ssh(ui, repo, node=None, **kwargs):
     extras = _extras_from_ui(ui)
     if extras.get('SSH'):
         return pre_push(ui, repo, node, **kwargs)
 
     return 0
 
 
 def pre_push_ssh_auth(ui, repo, node=None, **kwargs):
     """
     Mercurial pre_push hook for SSH
     """
     extras = _extras_from_ui(ui)
     if extras.get('SSH'):
         permission = extras['SSH_PERMISSIONS']
 
         if 'repository.write' == permission or 'repository.admin' == permission:
             return 0
 
         # non-zero ret code
         return 1
 
     return 0
 
 
 def post_push(ui, repo, node, **kwargs):
     """
     Mercurial post_push hook
     """
     extras = _extras_from_ui(ui)
 
     commit_ids = []
     branches = []
     bookmarks = []
     tags = []
 
     commits, _heads = _rev_range_hash(repo, node)
     for commit_id, branch in commits:
         commit_ids.append(commit_id)
         if branch not in branches:
             branches.append(branch)
 
     if hasattr(ui, '_rc_pushkey_branches'):
         bookmarks = ui._rc_pushkey_branches
 
     extras['hook_type'] = kwargs.get('hooktype', 'post_push')
     extras['commit_ids'] = commit_ids
     extras['new_refs'] = {
         'branches': branches,
         'bookmarks': bookmarks,
         'tags': tags
     }
 
     return _call_hook('post_push', extras, HgMessageWriter(ui))
 
 
 def post_push_ssh(ui, repo, node, **kwargs):
     """
     Mercurial post_push hook for SSH
     """
     if _extras_from_ui(ui).get('SSH'):
         return post_push(ui, repo, node, **kwargs)
     return 0
 
 
 def key_push(ui, repo, **kwargs):
     from vcsserver.hgcompat import get_ctx
     if kwargs['new'] != '0' and kwargs['namespace'] == 'bookmarks':
         # store new bookmarks in our UI object propagated later to post_push
         ui._rc_pushkey_branches = get_ctx(repo, kwargs['key']).bookmarks()
     return
 
 
 # backward compat
 log_pull_action = post_pull
 
 # backward compat
 log_push_action = post_push
 
 
 def handle_git_pre_receive(unused_repo_path, unused_revs, unused_env):
     """
     Old hook name: keep here for backward compatibility.
 
     This is only required when the installed git hooks are not upgraded.
     """
     pass
 
 
 def handle_git_post_receive(unused_repo_path, unused_revs, unused_env):
     """
     Old hook name: keep here for backward compatibility.
 
     This is only required when the installed git hooks are not upgraded.
     """
     pass
 
 
 HookResponse = collections.namedtuple('HookResponse', ('status', 'output'))
 
 
 def git_pre_pull(extras):
     """
     Pre pull hook.
 
     :param extras: dictionary containing the keys defined in simplevcs
     :type extras: dict
 
     :return: status code of the hook. 0 for success.
     :rtype: int
     """
 
     if 'pull' not in extras['hooks']:
         return HookResponse(0, '')
 
     stdout = io.BytesIO()
     try:
         status = _call_hook('pre_pull', extras, GitMessageWriter(stdout))
 
     except Exception as error:
         log.exception('Failed to call pre_pull hook')
         status = 128
         stdout.write(safe_bytes(f'ERROR: {error}\n'))
 
     return HookResponse(status, stdout.getvalue())
 
 
 def git_post_pull(extras):
     """
     Post pull hook.
 
     :param extras: dictionary containing the keys defined in simplevcs
     :type extras: dict
 
     :return: status code of the hook. 0 for success.
     :rtype: int
     """
     if 'pull' not in extras['hooks']:
         return HookResponse(0, '')
 
     stdout = io.BytesIO()
     try:
         status = _call_hook('post_pull', extras, GitMessageWriter(stdout))
     except Exception as error:
         status = 128
         stdout.write(safe_bytes(f'ERROR: {error}\n'))
 
     return HookResponse(status, stdout.getvalue())
 
 
 def _parse_git_ref_lines(revision_lines):
     rev_data = []
     for revision_line in revision_lines or []:
         old_rev, new_rev, ref = revision_line.strip().split(' ')
         ref_data = ref.split('/', 2)
         if ref_data[1] in ('tags', 'heads'):
             rev_data.append({
                 # NOTE(marcink):
                 # we're unable to tell total_commits for git at this point
                 # but we set the variable for consistency with GIT
                 'total_commits': -1,
                 'old_rev': old_rev,
                 'new_rev': new_rev,
                 'ref': ref,
                 'type': ref_data[1],
                 'name': ref_data[2],
             })
     return rev_data
 
 
 def git_pre_receive(unused_repo_path, revision_lines, env):
     """
     Pre push hook.
 
     :param extras: dictionary containing the keys defined in simplevcs
     :type extras: dict
 
     :return: status code of the hook. 0 for success.
     :rtype: int
     """
     extras = json.loads(env['RC_SCM_DATA'])
     rev_data = _parse_git_ref_lines(revision_lines)
     if 'push' not in extras['hooks']:
         return 0
     empty_commit_id = '0' * 40
 
     detect_force_push = extras.get('detect_force_push')
 
     for push_ref in rev_data:
         # store our git-env which holds the temp store
         push_ref['git_env'] = _get_git_env()
         push_ref['pruned_sha'] = ''
         if not detect_force_push:
             # don't check for forced-push when we don't need to
             continue
 
         type_ = push_ref['type']
         new_branch = push_ref['old_rev'] == empty_commit_id
         delete_branch = push_ref['new_rev'] == empty_commit_id
         if type_ == 'heads' and not (new_branch or delete_branch):
             old_rev = push_ref['old_rev']
             new_rev = push_ref['new_rev']
             cmd = [settings.GIT_EXECUTABLE, 'rev-list', old_rev, '^{}'.format(new_rev)]
             stdout, stderr = subprocessio.run_command(
                 cmd, env=os.environ.copy())
             # means we're having some non-reachable objects, this forced push was used
             if stdout:
                 push_ref['pruned_sha'] = stdout.splitlines()
 
     extras['hook_type'] = 'pre_receive'
     extras['commit_ids'] = rev_data
     return _call_hook('pre_push', extras, GitMessageWriter())
 
 
 def git_post_receive(unused_repo_path, revision_lines, env):
     """
     Post push hook.
 
     :param extras: dictionary containing the keys defined in simplevcs
     :type extras: dict
 
     :return: status code of the hook. 0 for success.
     :rtype: int
     """
     extras = json.loads(env['RC_SCM_DATA'])
     if 'push' not in extras['hooks']:
         return 0
 
     rev_data = _parse_git_ref_lines(revision_lines)
 
     git_revs = []
 
     # N.B.(skreft): it is ok to just call git, as git before calling a
     # subcommand sets the PATH environment variable so that it point to the
     # correct version of the git executable.
     empty_commit_id = '0' * 40
     branches = []
     tags = []
     for push_ref in rev_data:
         type_ = push_ref['type']
 
         if type_ == 'heads':
             if push_ref['old_rev'] == empty_commit_id:
                 # starting new branch case
                 if push_ref['name'] not in branches:
                     branches.append(push_ref['name'])
 
                 # Fix up head revision if needed
                 cmd = [settings.GIT_EXECUTABLE, 'show', 'HEAD']
                 try:
                     subprocessio.run_command(cmd, env=os.environ.copy())
                 except Exception:
                     push_ref_name = push_ref['name']
                     cmd = [settings.GIT_EXECUTABLE, 'symbolic-ref', '"HEAD"', f'"refs/heads/{push_ref_name}"']
                     print(f"Setting default branch to {push_ref_name}")
                     subprocessio.run_command(cmd, env=os.environ.copy())
 
                 cmd = [settings.GIT_EXECUTABLE, 'for-each-ref',
                        '--format=%(refname)', 'refs/heads/*']
                 stdout, stderr = subprocessio.run_command(
                     cmd, env=os.environ.copy())
                 heads = stdout
                 heads = heads.replace(push_ref['ref'], '')
                 heads = ' '.join(head for head
                                  in heads.splitlines() if head) or '.'
                 cmd = [settings.GIT_EXECUTABLE, 'log', '--reverse',
                        '--pretty=format:%H', '--', push_ref['new_rev'],
                        '--not', heads]
                 stdout, stderr = subprocessio.run_command(
                     cmd, env=os.environ.copy())
                 git_revs.extend(stdout.splitlines())
             elif push_ref['new_rev'] == empty_commit_id:
                 # delete branch case
                 git_revs.append('delete_branch=>%s' % push_ref['name'])
             else:
                 if push_ref['name'] not in branches:
                     branches.append(push_ref['name'])
 
                 cmd = [settings.GIT_EXECUTABLE, 'log',
                        '{old_rev}..{new_rev}'.format(**push_ref),
                        '--reverse', '--pretty=format:%H']
                 stdout, stderr = subprocessio.run_command(
                     cmd, env=os.environ.copy())
                 git_revs.extend(stdout.splitlines())
         elif type_ == 'tags':
             if push_ref['name'] not in tags:
                 tags.append(push_ref['name'])
             git_revs.append('tag=>%s' % push_ref['name'])
 
     extras['hook_type'] = 'post_receive'
     extras['commit_ids'] = git_revs
     extras['new_refs'] = {
         'branches': branches,
         'bookmarks': [],
         'tags': tags,
     }
 
     if 'repo_size' in extras['hooks']:
         try:
             _call_hook('repo_size', extras, GitMessageWriter())
-        except:
+        except Exception:
             pass
 
     return _call_hook('post_push', extras, GitMessageWriter())
 
 
 def _get_extras_from_txn_id(path, txn_id):
     extras = {}
     try:
         cmd = [settings.SVNLOOK_EXECUTABLE, 'pget',
                '-t', txn_id,
                '--revprop', path, 'rc-scm-extras']
         stdout, stderr = subprocessio.run_command(
             cmd, env=os.environ.copy())
         extras = json.loads(base64.urlsafe_b64decode(stdout))
     except Exception:
         log.exception('Failed to extract extras info from txn_id')
 
     return extras
 
 
 def _get_extras_from_commit_id(commit_id, path):
     extras = {}
     try:
         cmd = [settings.SVNLOOK_EXECUTABLE, 'pget',
                '-r', commit_id,
                '--revprop', path, 'rc-scm-extras']
         stdout, stderr = subprocessio.run_command(
             cmd, env=os.environ.copy())
         extras = json.loads(base64.urlsafe_b64decode(stdout))
     except Exception:
         log.exception('Failed to extract extras info from commit_id')
 
     return extras
 
 
 def svn_pre_commit(repo_path, commit_data, env):
     path, txn_id = commit_data
     branches = []
     tags = []
 
     if env.get('RC_SCM_DATA'):
         extras = json.loads(env['RC_SCM_DATA'])
     else:
         # fallback method to read from TXN-ID stored data
         extras = _get_extras_from_txn_id(path, txn_id)
         if not extras:
             return 0
 
     extras['hook_type'] = 'pre_commit'
     extras['commit_ids'] = [txn_id]
     extras['txn_id'] = txn_id
     extras['new_refs'] = {
         'total_commits': 1,
         'branches': branches,
         'bookmarks': [],
         'tags': tags,
     }
 
     return _call_hook('pre_push', extras, SvnMessageWriter())
 
 
 def svn_post_commit(repo_path, commit_data, env):
     """
     commit_data is path, rev, txn_id
     """
     if len(commit_data) == 3:
         path, commit_id, txn_id = commit_data
     elif len(commit_data) == 2:
         log.error('Failed to extract txn_id from commit_data using legacy method. '
                   'Some functionality might be limited')
         path, commit_id = commit_data
         txn_id = None
 
     branches = []
     tags = []
 
     if env.get('RC_SCM_DATA'):
         extras = json.loads(env['RC_SCM_DATA'])
     else:
         # fallback method to read from TXN-ID stored data
         extras = _get_extras_from_commit_id(commit_id, path)
         if not extras:
             return 0
 
     extras['hook_type'] = 'post_commit'
     extras['commit_ids'] = [commit_id]
     extras['txn_id'] = txn_id
     extras['new_refs'] = {
         'branches': branches,
         'bookmarks': [],
         'tags': tags,
         'total_commits': 1,
     }
 
     if 'repo_size' in extras['hooks']:
         try:
             _call_hook('repo_size', extras, SvnMessageWriter())
         except Exception:
             pass
 
     return _call_hook('post_push', extras, SvnMessageWriter())

vcsserver/subprocessio.py

@@ -1,563 +1,563 b''
 """
 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[at]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 collections
 import logging
 import subprocess
 import threading
 
 from vcsserver.str_utils import safe_str
 
 log = logging.getLogger(__name__)
 
 
 class StreamFeeder(threading.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 isinstance(source, int):  # file pointer it is
                 # converting file descriptor (int) stdin into file-like
                 source = os.fdopen(source, 'rb', 16384)
             # let's see if source is file-like by now
             filelike = hasattr(source, 'read')
         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):
         writer = self.writeiface
         try:
             if self.bytes:
                 os.write(writer, self.bytes)
             else:
                 s = self.source
 
                 while 1:
                     _bytes = s.read(4096)
                     if not _bytes:
                         break
                     os.write(writer, _bytes)
 
         finally:
             os.close(writer)
 
     @property
     def output(self):
         return self.readiface
 
 
 class InputStreamChunker(threading.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 = threading.Event()
         self.data_added.clear()
 
         self.keep_reading = threading.Event()
         self.keep_reading.set()
 
         self.EOF = threading.Event()
         self.EOF.clear()
 
         self.go = threading.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:
+        except Exception:
             pass
 
     def run(self):
         s = self.source
         t = self.target
         cs = self.chunk_size
         chunk_count_max = self.chunk_count_max
         keep_reading = self.keep_reading
         da = self.data_added
         go = self.go
 
         try:
             b = s.read(cs)
         except ValueError:
             b = ''
 
         timeout_input = 20
         while b and go.is_set():
             if len(t) > chunk_count_max:
                 keep_reading.clear()
                 keep_reading.wait(timeout_input)
                 if len(t) > chunk_count_max + timeout_input:
                     log.error("Timed out while waiting for input from subprocess.")
                     os._exit(-1)  # this will cause the worker to recycle itself
 
             t.append(b)
             da.set()
 
             try:
                 b = s.read(cs)
             except ValueError:  # probably "I/O operation on closed file"
                 b = ''
 
         self.EOF.set()
         da.set()  # for cases when done but there was no input.
 
 
 class BufferedGenerator(object):
     """
     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, name, source, buffer_size=65536, chunk_size=4096,
                  starting_values=None, bottomless=False):
         starting_values = starting_values or []
         self.name = name
         self.buffer_size = buffer_size
         self.chunk_size = chunk_size
 
         if bottomless:
             maxlen = int(buffer_size / chunk_size)
         else:
             maxlen = None
 
         self.data_queue = collections.deque(starting_values, maxlen)
         self.worker = InputStreamChunker(source, self.data_queue, buffer_size, chunk_size)
         if starting_values:
             self.worker.data_added.set()
         self.worker.start()
 
     ####################
     # Generator's methods
     ####################
     def __str__(self):
         return f'BufferedGenerator(name={self.name} chunk: {self.chunk_size} on buffer: {self.buffer_size})'
 
     def __iter__(self):
         return self
 
     def __next__(self):
 
         while not self.length and not self.worker.EOF.is_set():
             self.worker.data_added.clear()
             self.worker.data_added.wait(0.2)
 
         if self.length:
             self.worker.keep_reading.set()
             return bytes(self.data_queue.popleft())
         elif self.worker.EOF.is_set():
             raise StopIteration
 
     def throw(self, exc_type, value=None, traceback=None):
         if not self.worker.EOF.is_set():
             raise exc_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
 
     ####################
     # 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.
 
         :returns: An Event class instance.
         """
         return self.worker.EOF
 
     @property
     def done_reading(self):
         """
         Done_reading 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.
 
         :returns: An Bool value.
         """
         return self.worker.EOF.is_set()
 
     @property
     def length(self):
         """
         returns int.
 
         This is the length of the queue of chunks, not the length of
         the combined contents in those chunks.
 
         __len__() cannot be meaningfully implemented because this
         reader is just flying through a bottomless pit content and
         can only know the length of what it already saw.
 
         If __len__() on WSGI server per PEP 3333 returns a value,
         the response'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_queue)
 
     def prepend(self, x):
         self.data_queue.appendleft(x)
 
     def append(self, x):
         self.data_queue.append(x)
 
     def extend(self, o):
         self.data_queue.extend(o)
 
     def __getitem__(self, i):
         return self.data_queue[i]
 
 
 class SubprocessIOChunker(object):
     """
     Processor class wrapping handling of subprocess IO.
 
     .. important::
 
        Watch out for the method `__del__` on this class. If this object
        is deleted, it will kill the subprocess, so avoid to
        return the `output` attribute or usage of it like in the following
        example::
 
           # `args` expected to run a program that produces a lot of output
           output = ''.join(SubprocessIOChunker(
              args, shell=False, inputstream=inputstream, env=environ).output)
 
           # `output` will not contain all the data, because the __del__ method
           # has already killed the subprocess in this case before all output
           # has been consumed.
 
 
 
     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 str)
       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
     an iterable 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 occurred 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
     OSError family of exceptions
 
     Example usage:
     #    try:
     #        answer = SubprocessIOChunker(
     #            cmd,
     #            input,
     #            buffer_size = 65536,
     #            chunk_size = 4096
     #            )
     #    except (OSError) as e:
     #        print str(e)
     #        raise e
     #
     #    return answer
 
 
     """
 
     # TODO: johbo: This is used to make sure that the open end of the PIPE
     # is closed in the end. It would be way better to wrap this into an
     # object, so that it is closed automatically once it is consumed or
     # something similar.
     _close_input_fd = None
 
     _closed = False
     _stdout = None
     _stderr = None
 
     def __init__(self, cmd, input_stream=None, buffer_size=65536,
                  chunk_size=4096, starting_values=None, fail_on_stderr=True,
                  fail_on_return_code=True, **kwargs):
         """
         Initializes SubprocessIOChunker
 
         :param cmd: A Subprocess.Popen style "cmd". Can be string or array of strings
         :param input_stream: (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.
         :param fail_on_stderr: (Default: True) Whether to raise an exception in
                                case something is written to stderr.
         :param fail_on_return_code: (Default: True) Whether to raise an
                                     exception if the return code is not 0.
         """
 
         kwargs['shell'] = kwargs.get('shell', True)
 
         starting_values = starting_values or []
         if input_stream:
             input_streamer = StreamFeeder(input_stream)
             input_streamer.start()
             input_stream = input_streamer.output
             self._close_input_fd = input_stream
 
         self._fail_on_stderr = fail_on_stderr
         self._fail_on_return_code = fail_on_return_code
         self.cmd = cmd
 
         _p = subprocess.Popen(cmd, bufsize=-1, stdin=input_stream, stdout=subprocess.PIPE, stderr=subprocess.PIPE,
                               **kwargs)
         self.process = _p
 
         bg_out = BufferedGenerator('stdout', _p.stdout, buffer_size, chunk_size, starting_values)
         bg_err = BufferedGenerator('stderr', _p.stderr, 10240, 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(0.2)
             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.
         return_code = _p.poll()
         ret_code_ok = return_code in [None, 0]
         ret_code_fail = return_code is not None and return_code != 0
         if (
             (ret_code_fail and fail_on_return_code) or
             (ret_code_ok and fail_on_stderr and bg_err.length)
         ):
 
             try:
                 _p.terminate()
             except Exception:
                 pass
 
             bg_out.stop()
             out = b''.join(bg_out)
             self._stdout = out
 
             bg_err.stop()
             err = b''.join(bg_err)
             self._stderr = err
 
             # code from https://github.com/schacon/grack/pull/7
             if err.strip() == b'fatal: The remote end hung up unexpectedly' and out.startswith(b'0034shallow '):
                 bg_out = iter([out])
                 _p = None
             elif err and fail_on_stderr:
                 text_err = err.decode()
                 raise OSError(
                     "Subprocess exited due to an error:\n{}".format(text_err))
 
             if ret_code_fail and fail_on_return_code:
                 text_err = err.decode()
                 if not err:
                     # maybe get empty stderr, try stdout instead
                     # in many cases git reports the errors on stdout too
                     text_err = out.decode()
                 raise OSError(
                     "Subprocess exited with non 0 ret code:{}: stderr:{}".format(return_code, text_err))
 
         self.stdout = bg_out
         self.stderr = bg_err
         self.inputstream = input_stream
 
     def __str__(self):
         proc = getattr(self, 'process', 'NO_PROCESS')
         return f'SubprocessIOChunker: {proc}'
 
     def __iter__(self):
         return self
 
     def __next__(self):
         # Note: mikhail: We need to be sure that we are checking the return
         # code after the stdout stream is closed. Some processes, e.g. git
         # are doing some magic in between closing stdout and terminating the
         # process and, as a result,  we are not getting return code on "slow"
         # systems.
         result = None
         stop_iteration = None
         try:
             result = next(self.stdout)
         except StopIteration as e:
             stop_iteration = e
 
         if self.process:
             return_code = self.process.poll()
             ret_code_fail = return_code is not None and return_code != 0
             if ret_code_fail and self._fail_on_return_code:
                 self.stop_streams()
                 err = self.get_stderr()
                 raise OSError(
                     "Subprocess exited (exit_code:{}) due to an error during iteration:\n{}".format(return_code, err))
 
         if stop_iteration:
             raise stop_iteration
         return result
 
     def throw(self, exc_type, value=None, traceback=None):
         if self.stdout.length or not self.stdout.done_reading:
             raise exc_type(value)
 
     def close(self):
         if self._closed:
             return
 
         try:
             self.process.terminate()
         except Exception:
             pass
         if self._close_input_fd:
             os.close(self._close_input_fd)
         try:
             self.stdout.close()
         except Exception:
             pass
         try:
             self.stderr.close()
         except Exception:
             pass
         try:
             os.close(self.inputstream)
         except Exception:
             pass
 
         self._closed = True
 
     def stop_streams(self):
         getattr(self.stdout, 'stop', lambda: None)()
         getattr(self.stderr, 'stop', lambda: None)()
 
     def get_stdout(self):
         if self._stdout:
             return self._stdout
         else:
             return b''.join(self.stdout)
 
     def get_stderr(self):
         if self._stderr:
             return self._stderr
         else:
             return b''.join(self.stderr)
 
 
 def run_command(arguments, env=None):
     """
     Run the specified command and return the stdout.
 
     :param arguments: sequence of program arguments (including the program name)
     :type arguments: list[str]
     """
 
     cmd = arguments
     log.debug('Running subprocessio command %s', cmd)
     proc = None
     try:
         _opts = {'shell': False, 'fail_on_stderr': False}
         if env:
             _opts.update({'env': env})
         proc = SubprocessIOChunker(cmd, **_opts)
         return b''.join(proc), b''.join(proc.stderr)
     except OSError as err:
         cmd = ' '.join(map(safe_str, cmd)) # human friendly CMD
         tb_err = ("Couldn't run subprocessio command (%s).\n"
                   "Original error was:%s\n" % (cmd, err))
         log.exception(tb_err)
         raise Exception(tb_err)
     finally:
         if proc:
             proc.close()
 

vcsserver/tweens/request_wrapper.py

@@ -1,114 +1,112 b''
 # RhodeCode VCSServer provides access to different vcs backends via network.
 # Copyright (C) 2014-2020 RhodeCode GmbH
 #
 # 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 3 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301  USA
 
 import time
 import logging
 
-import msgpack
-
 import vcsserver
 from vcsserver.str_utils import safe_str, ascii_str
 
 log = logging.getLogger(__name__)
 
 
 def get_access_path(environ):
     path = environ.get('PATH_INFO')
     return path
 
 
 def get_user_agent(environ):
     return environ.get('HTTP_USER_AGENT')
 
 
 def get_call_context(registry) -> dict:
     cc = {}
     if hasattr(registry, 'vcs_call_context'):
         cc.update({
             'X-RC-Method': registry.vcs_call_context.get('method'),
             'X-RC-Repo-Name': registry.vcs_call_context.get('repo_name')
         })
 
     return cc
 
 
 class RequestWrapperTween(object):
     def __init__(self, handler, registry):
         self.handler = handler
         self.registry = registry
 
         # one-time configuration code goes here
 
     def __call__(self, request):
         start = time.time()
         log.debug('Starting request time measurement')
         response = None
 
         try:
             response = self.handler(request)
         finally:
             ua = get_user_agent(request.environ)
             call_context = get_call_context(request.registry)
             vcs_method = call_context.get('X-RC-Method', '_NO_VCS_METHOD')
             repo_name = call_context.get('X-RC-Repo-Name', '')
 
             count = request.request_count()
             _ver_ = ascii_str(vcsserver.__version__)
             _path = safe_str(get_access_path(request.environ))
 
             ip = '127.0.0.1'
             match_route = request.matched_route.name if request.matched_route else "NOT_FOUND"
             resp_code = getattr(response, 'status_code', 'UNDEFINED')
 
             _view_path = f"{repo_name}@{_path}/{vcs_method}"
 
             total = time.time() - start
 
             log.info(
                 'Req[%4s] IP: %s %s Request to %s time: %.4fs [%s], VCSServer %s',
                 count, ip, request.environ.get('REQUEST_METHOD'),
                 _view_path, total, ua, _ver_,
                 extra={"time": total, "ver": _ver_, "code": resp_code,
                        "path": _path, "view_name": match_route, "user_agent": ua,
                        "vcs_method": vcs_method, "repo_name": repo_name}
             )
 
             statsd = request.registry.statsd
             if statsd:
                 match_route = request.matched_route.name if request.matched_route else _path
                 elapsed_time_ms = round(1000.0 * total)  # use ms only
                 statsd.timing(
                     "vcsserver_req_timing.histogram", elapsed_time_ms,
                     tags=[
                         "view_name:{}".format(match_route),
                         "code:{}".format(resp_code)
                     ],
                     use_decimals=False
                 )
                 statsd.incr(
                     "vcsserver_req_total", tags=[
                         "view_name:{}".format(match_route),
                         "code:{}".format(resp_code)
                     ])
 
         return response
 
 
 def includeme(config):
     config.add_tween(
         'vcsserver.tweens.request_wrapper.RequestWrapperTween',
     )