[MERGE] merged trunk.
authorVo Minh Thu <vmt@openerp.com>
Mon, 13 Feb 2012 11:12:33 +0000 (12:12 +0100)
committerVo Minh Thu <vmt@openerp.com>
Mon, 13 Feb 2012 11:12:33 +0000 (12:12 +0100)
bzr revid: vmt@openerp.com-20120213111233-ontycsea5cpnvsgv

1  2 
gunicorn.conf.py
openerp/addons/base/ir/ir_ui_menu.py
openerp/service/__init__.py
openerp/wsgi/core.py

@@@ -17,12 -21,13 +21,12 @@@ pidfile = '.gunicorn.pid
  # Gunicorn recommends 2-4 x number_of_cpu_cores, but
  # you'll want to vary this a bit to find the best for your
  # particular work load.
 -workers = 4
 +workers = 10
  
  # Some application-wide initialization is needed.
- on_starting = openerp.wsgi.on_starting
- pre_request = openerp.wsgi.pre_request
- post_request = openerp.wsgi.post_request
+ on_starting = openerp.wsgi.core.on_starting
 -when_ready = openerp.wsgi.core.when_ready
+ pre_request = openerp.wsgi.core.pre_request
+ post_request = openerp.wsgi.core.post_request
  
  # openerp request-response cycle can be quite long for
  # big reports for example
Simple merge
Simple merge
index 0000000,9862225..379ccaf
mode 000000,100644..100644
--- /dev/null
@@@ -1,0 -1,543 +1,517 @@@
+ # -*- coding: utf-8 -*-
+ ##############################################################################
+ #
+ #    OpenERP, Open Source Management Solution
+ #    Copyright (C) 2011-2012 OpenERP s.a. (<http://openerp.com>).
+ #
+ #    This program is free software: you can redistribute it and/or modify
+ #    it under the terms of the GNU Affero 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 Affero General Public License for more details.
+ #
+ #    You should have received a copy of the GNU Affero General Public License
+ #    along with this program.  If not, see <http://www.gnu.org/licenses/>.
+ #
+ ##############################################################################
+ """
+ WSGI stack, common code.
+ """
+ import httplib
+ import urllib
+ import xmlrpclib
+ import StringIO
+ import errno
+ import logging
+ import os
+ import signal
+ import sys
+ import threading
+ import traceback
+ import openerp
+ import openerp.modules
+ import openerp.tools.config as config
+ from ..service import websrv_lib
+ _logger = logging.getLogger(__name__)
+ # XML-RPC fault codes. Some care must be taken when changing these: the
+ # constants are also defined client-side and must remain in sync.
+ # User code must use the exceptions defined in ``openerp.exceptions`` (not
+ # create directly ``xmlrpclib.Fault`` objects).
+ RPC_FAULT_CODE_CLIENT_ERROR = 1 # indistinguishable from app. error.
+ RPC_FAULT_CODE_APPLICATION_ERROR = 1
+ RPC_FAULT_CODE_WARNING = 2
+ RPC_FAULT_CODE_ACCESS_DENIED = 3
+ RPC_FAULT_CODE_ACCESS_ERROR = 4
+ # The new (6.1) versioned RPC paths.
+ XML_RPC_PATH = '/openerp/xmlrpc'
+ XML_RPC_PATH_1 = '/openerp/xmlrpc/1'
+ JSON_RPC_PATH = '/openerp/jsonrpc'
+ JSON_RPC_PATH_1 = '/openerp/jsonrpc/1'
+ def xmlrpc_return(start_response, service, method, params, legacy_exceptions=False):
+     """
+     Helper to call a service's method with some params, using a wsgi-supplied
+     ``start_response`` callback.
+     This is the place to look at to see the mapping between core exceptions
+     and XML-RPC fault codes.
+     """
+     # Map OpenERP core exceptions to XML-RPC fault codes. Specific exceptions
+     # defined in ``openerp.exceptions`` are mapped to specific fault codes;
+     # all the other exceptions are mapped to the generic
+     # RPC_FAULT_CODE_APPLICATION_ERROR value.
+     # This also mimics SimpleXMLRPCDispatcher._marshaled_dispatch() for
+     # exception handling.
+     try:
+         result = openerp.netsvc.dispatch_rpc(service, method, params)
+         response = xmlrpclib.dumps((result,), methodresponse=1, allow_none=False, encoding=None)
+     except Exception, e:
+         if legacy_exceptions:
+             response = xmlrpc_handle_exception_legacy(e)
+         else:
+             response = xmlrpc_handle_exception(e)
+     start_response("200 OK", [('Content-Type','text/xml'), ('Content-Length', str(len(response)))])
+     return [response]
+ def xmlrpc_handle_exception(e):
+     if isinstance(e, openerp.osv.osv.except_osv): # legacy
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_WARNING, openerp.tools.ustr(e.value))
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.Warning):
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_WARNING, str(e))
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance (e, openerp.exceptions.AccessError):
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_ACCESS_ERROR, str(e))
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.AccessDenied):
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_ACCESS_DENIED, str(e))
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.DeferredException):
+         info = e.traceback
+         # Which one is the best ?
+         formatted_info = "".join(traceback.format_exception(*info))
+         #formatted_info = openerp.tools.exception_to_unicode(e) + '\n' + info
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_APPLICATION_ERROR, formatted_info)
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     else:
+         if hasattr(e, 'message') and e.message == 'AccessDenied': # legacy
+             fault = xmlrpclib.Fault(RPC_FAULT_CODE_ACCESS_DENIED, str(e))
+             response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+         else:
+             info = sys.exc_info()
+             # Which one is the best ?
+             formatted_info = "".join(traceback.format_exception(*info))
+             #formatted_info = openerp.tools.exception_to_unicode(e) + '\n' + info
+             fault = xmlrpclib.Fault(RPC_FAULT_CODE_APPLICATION_ERROR, formatted_info)
+             response = xmlrpclib.dumps(fault, allow_none=None, encoding=None)
+     return response
+ def xmlrpc_handle_exception_legacy(e):
+     if isinstance(e, openerp.osv.osv.except_osv):
+         fault = xmlrpclib.Fault('warning -- ' + e.name + '\n\n' + e.value, '')
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.Warning):
+         fault = xmlrpclib.Fault('warning -- Warning\n\n' + str(e), '')
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.AccessError):
+         fault = xmlrpclib.Fault('warning -- AccessError\n\n' + str(e), '')
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.AccessDenied):
+         fault = xmlrpclib.Fault('AccessDenied', str(e))
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     elif isinstance(e, openerp.exceptions.DeferredException):
+         info = e.traceback
+         formatted_info = "".join(traceback.format_exception(*info))
+         fault = xmlrpclib.Fault(openerp.tools.ustr(e.message), formatted_info)
+         response = xmlrpclib.dumps(fault, allow_none=False, encoding=None)
+     else:
+         info = sys.exc_info()
+         formatted_info = "".join(traceback.format_exception(*info))
+         fault = xmlrpclib.Fault(openerp.tools.exception_to_unicode(e), formatted_info)
+         response = xmlrpclib.dumps(fault, allow_none=None, encoding=None)
+     return response
+ def wsgi_xmlrpc_1(environ, start_response):
+     """ The main OpenERP WSGI handler."""
+     if environ['REQUEST_METHOD'] == 'POST' and environ['PATH_INFO'].startswith(XML_RPC_PATH_1):
+         length = int(environ['CONTENT_LENGTH'])
+         data = environ['wsgi.input'].read(length)
+         params, method = xmlrpclib.loads(data)
+         path = environ['PATH_INFO'][len(XML_RPC_PATH_1):]
+         if path.startswith('/'): path = path[1:]
+         if path.endswith('/'): path = path[:-1]
+         path = path.split('/')
+         # All routes are hard-coded.
+         # No need for a db segment.
+         if len(path) == 1:
+             service = path[0]
+             if service == 'common':
+                 if method in ('server_version',):
+                     service = 'db'
+             return xmlrpc_return(start_response, service, method, params)
+         # A db segment must be given.
+         elif len(path) == 2:
+             service, db_name = path
+             params = (db_name,) + params
+             return xmlrpc_return(start_response, service, method, params)
+         # A db segment and a model segment must be given.
+         elif len(path) == 3 and path[0] == 'model':
+             service, db_name, model_name = path
+             params = (db_name,) + params[:2] + (model_name,) + params[2:]
+             service = 'object'
+             return xmlrpc_return(start_response, service, method, params)
+         # The body has been read, need to raise an exception (not return None).
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_CLIENT_ERROR, '')
+         response = xmlrpclib.dumps(fault, allow_none=None, encoding=None)
+         start_response("200 OK", [('Content-Type','text/xml'), ('Content-Length', str(len(response)))])
+         return [response]
+ def wsgi_xmlrpc(environ, start_response):
+     """ WSGI handler to return the versions."""
+     if environ['REQUEST_METHOD'] == 'POST' and environ['PATH_INFO'].startswith(XML_RPC_PATH):
+         length = int(environ['CONTENT_LENGTH'])
+         data = environ['wsgi.input'].read(length)
+         params, method = xmlrpclib.loads(data)
+         path = environ['PATH_INFO'][len(XML_RPC_PATH):]
+         if path.startswith('/'): path = path[1:]
+         if path.endswith('/'): path = path[:-1]
+         path = path.split('/')
+         # All routes are hard-coded.
+         if len(path) == 1 and path[0] == '' and method in ('version',):
+             return xmlrpc_return(start_response, 'common', method, ())
+         # The body has been read, need to raise an exception (not return None).
+         fault = xmlrpclib.Fault(RPC_FAULT_CODE_CLIENT_ERROR, '')
+         response = xmlrpclib.dumps(fault, allow_none=None, encoding=None)
+         start_response("200 OK", [('Content-Type','text/xml'), ('Content-Length', str(len(response)))])
+         return [response]
+ def wsgi_xmlrpc_legacy(environ, start_response):
+     if environ['REQUEST_METHOD'] == 'POST' and environ['PATH_INFO'].startswith('/xmlrpc/'):
+         length = int(environ['CONTENT_LENGTH'])
+         data = environ['wsgi.input'].read(length)
+         path = environ['PATH_INFO'][len('/xmlrpc/'):] # expected to be one of db, object, ...
+         params, method = xmlrpclib.loads(data)
+         return xmlrpc_return(start_response, path, method, params, True)
+ def wsgi_jsonrpc(environ, start_response):
+     pass
+ def wsgi_webdav(environ, start_response):
+     pi = environ['PATH_INFO']
+     if environ['REQUEST_METHOD'] == 'OPTIONS' and pi in ['*','/']:
+         return return_options(environ, start_response)
+     elif pi.startswith('/webdav'):
+         http_dir = websrv_lib.find_http_service(pi)
+         if http_dir:
+             path = pi[len(http_dir.path):]
+             if path.startswith('/'):
+                 environ['PATH_INFO'] = path
+             else:
+                 environ['PATH_INFO'] = '/' + path
+             return http_to_wsgi(http_dir)(environ, start_response)
+ def return_options(environ, start_response):
+     # Microsoft specific header, see
+     # http://www.ibm.com/developerworks/rational/library/2089.html
+     if 'Microsoft' in environ.get('User-Agent', ''):
+         options = [('MS-Author-Via', 'DAV')]
+     else:
+         options = []
+     options += [('DAV', '1 2'), ('Allow', 'GET HEAD PROPFIND OPTIONS REPORT')]
+     start_response("200 OK", [('Content-Length', str(0))] + options)
+     return []
+ def http_to_wsgi(http_dir):
+     """
+     Turn a BaseHTTPRequestHandler into a WSGI entry point.
+     Actually the argument is not a bare BaseHTTPRequestHandler but is wrapped
+     (as a class, so it needs to be instanciated) in a HTTPDir.
+     This code is adapted from wbsrv_lib.MultiHTTPHandler._handle_one_foreign().
+     It is a temporary solution: the HTTP sub-handlers (in particular the
+     document_webdav addon) have to be WSGIfied.
+     """
+     def wsgi_handler(environ, start_response):
+         headers = {}
+         for key, value in environ.items():
+             if key.startswith('HTTP_'):
+                 key = key[5:].replace('_', '-').title()
+                 headers[key] = value
+             if key == 'CONTENT_LENGTH':
+                 key = key.replace('_', '-').title()
+                 headers[key] = value
+         if environ.get('Content-Type'):
+             headers['Content-Type'] = environ['Content-Type']
+         path = urllib.quote(environ.get('PATH_INFO', ''))
+         if environ.get('QUERY_STRING'):
+             path += '?' + environ['QUERY_STRING']
+         request_version = 'HTTP/1.1' # TODO
+         request_line = "%s %s %s\n" % (environ['REQUEST_METHOD'], path, request_version)
+         class Dummy(object):
+             pass
+         # Let's pretend we have a server to hand to the handler.
+         server = Dummy()
+         server.server_name = environ['SERVER_NAME']
+         server.server_port = int(environ['SERVER_PORT'])
+         # Initialize the underlying handler and associated auth. provider.
+         con = openerp.service.websrv_lib.noconnection(environ['wsgi.input'])
+         handler = http_dir.instanciate_handler(con, environ['REMOTE_ADDR'], server)
+         # Populate the handler as if it is called by a regular HTTP server
+         # and the request is already parsed.
+         handler.wfile = StringIO.StringIO()
+         handler.rfile = environ['wsgi.input']
+         handler.headers = headers
+         handler.command = environ['REQUEST_METHOD']
+         handler.path = path
+         handler.request_version = request_version
+         handler.close_connection = 1
+         handler.raw_requestline = request_line
+         handler.requestline = request_line
+         # Handle authentication if there is an auth. provider associated to
+         # the handler.
+         if hasattr(handler, 'auth_provider'):
+             try:
+                 handler.auth_provider.checkRequest(handler, path)
+             except websrv_lib.AuthRequiredExc, ae:
+                 # Darwin 9.x.x webdav clients will report "HTTP/1.0" to us, while they support (and need) the
+                 # authorisation features of HTTP/1.1 
+                 if request_version != 'HTTP/1.1' and ('Darwin/9.' not in handler.headers.get('User-Agent', '')):
+                     start_response("403 Forbidden", [])
+                     return []
+                 start_response("401 Authorization required", [
+                     ('WWW-Authenticate', '%s realm="%s"' % (ae.atype,ae.realm)),
+                     # ('Connection', 'keep-alive'),
+                     ('Content-Type', 'text/html'),
+                     ('Content-Length', 4), # len(self.auth_required_msg)
+                     ])
+                 return ['Blah'] # self.auth_required_msg
+             except websrv_lib.AuthRejectedExc,e:
+                 start_response("403 %s" % (e.args[0],), [])
+                 return []
+         method_name = 'do_' + handler.command
+         # Support the OPTIONS method even when not provided directly by the
+         # handler. TODO I would prefer to remove it and fix the handler if
+         # needed.
+         if not hasattr(handler, method_name):
+             if handler.command == 'OPTIONS':
+                 return return_options(environ, start_response)
+             start_response("501 Unsupported method (%r)" % handler.command, [])
+             return []
+         # Finally, call the handler's method.
+         try:
+             method = getattr(handler, method_name)
+             method()
+             # The DAV handler buffers its output and provides a _flush()
+             # method.
+             getattr(handler, '_flush', lambda: None)()
+             response = parse_http_response(handler.wfile.getvalue())
+             response_headers = response.getheaders()
+             body = response.read()
+             start_response(str(response.status) + ' ' + response.reason, response_headers)
+             return [body]
+         except (websrv_lib.AuthRejectedExc, websrv_lib.AuthRequiredExc):
+             raise
+         except Exception, e:
+             start_response("500 Internal error", [])
+             return []
+     return wsgi_handler
+ def parse_http_response(s):
+     """ Turn a HTTP response string into a httplib.HTTPResponse object."""
+     class DummySocket(StringIO.StringIO):
+         """
+         This is used to provide a StringIO to httplib.HTTPResponse
+         which, instead of taking a file object, expects a socket and
+         uses its makefile() method.
+         """
+         def makefile(self, *args, **kw):
+             return self
+     response = httplib.HTTPResponse(DummySocket(s))
+     response.begin()
+     return response
+ # WSGI handlers registered through the register_wsgi_handler() function below.
+ module_handlers = []
+ def register_wsgi_handler(handler):
+     """ Register a WSGI handler.
+     Handlers are tried in the order they are added. We might provide a way to
+     register a handler for specific routes later.
+     """
+     module_handlers.append(handler)
+ def application(environ, start_response):
+     """ WSGI entry point."""
+     # Try all handlers until one returns some result (i.e. not None).
+     wsgi_handlers = [
+         wsgi_xmlrpc_1,
+         wsgi_xmlrpc,
+         wsgi_jsonrpc,
+         wsgi_xmlrpc_legacy,
+         wsgi_webdav
+         ] + module_handlers
+     for handler in wsgi_handlers:
+         result = handler(environ, start_response)
+         if result is None:
+             continue
+         return result
+     # We never returned from the loop.
+     response = 'No handler found.\n'
+     start_response('404 Not Found', [('Content-Type', 'text/plain'), ('Content-Length', str(len(response)))])
+     return [response]
+ # The WSGI server, started by start_server(), stopped by stop_server().
+ httpd = None
+ def serve():
+     """ Serve HTTP requests via werkzeug development server.
+     If werkzeug can not be imported, we fall back to wsgiref's simple_server.
+     Calling this function is blocking, you might want to call it in its own
+     thread.
+     """
+     global httpd
+     # TODO Change the xmlrpc_* options to http_*
+     interface = config['xmlrpc_interface'] or '0.0.0.0'
+     port = config['xmlrpc_port']
+     try:
+         import werkzeug.serving
+         if config['proxy_mode']:
+             from werkzeug.contrib.fixers import ProxyFix
+             app = ProxyFix(application)
+             suffix = ' (in proxy mode)'
+         else:
+             app = application
+             suffix = ''
+         httpd = werkzeug.serving.make_server(interface, port, application, threaded=True)
+         _logger.info('HTTP service (werkzeug) running on %s:%s%s', interface, port, suffix)
+     except ImportError:
+         import wsgiref.simple_server
+         _logger.warning('Werkzeug module unavailable, falling back to wsgiref.')
+         if config['proxy_mode']:
+             _logger.warning('Werkzeug module unavailable, not using proxy mode.')
+         httpd = wsgiref.simple_server.make_server(interface, port, application)
+         _logger.info('HTTP service (wsgiref) running on %s:%s', interface, port)
+     httpd.serve_forever()
+ def start_server():
+     """ Call serve() in its own thread.
+     The WSGI server can be shutdown with stop_server() below.
+     """
 -    threading.Thread(target=serve).start()
++    threading.Thread(name='WSGI server', target=serve).start()
+ def stop_server():
+     """ Initiate the shutdown of the WSGI server.
+     The server is supposed to have been started by start_server() above.
+     """
+     if httpd:
+         httpd.shutdown()
+ # Master process id, can be used for signaling.
+ arbiter_pid = None
+ # Application setup before we can spawn any worker process.
+ # This is suitable for e.g. gunicorn's on_starting hook.
+ def on_starting(server):
+     global arbiter_pid
+     arbiter_pid = os.getpid() # TODO check if this is true even after replacing the executable
 -    #openerp.tools.cache = kill_workers_cache
++    openerp.multi_process = True # Yay!
+     openerp.netsvc.init_logger()
+     openerp.osv.osv.start_object_proxy()
+     openerp.service.web_services.start_web_services()
+     openerp.modules.module.initialize_sys_path()
+     openerp.modules.loading.open_openerp_namespace()
+     for m in openerp.conf.server_wide_modules:
+         try:
+             openerp.modules.module.load_openerp_module(m)
+         except Exception:
+             msg = ''
+             if m == 'web':
+                 msg = """
+ The `web` module is provided by the addons found in the `openerp-web` project.
+ Maybe you forgot to add those addons in your addons_path configuration."""
+             _logger.exception('Failed to load server-wide module `%s`.%s', m, msg)
 -# Install our own signal handler on the master process.
 -def when_ready(server):
 -    # Hijack gunicorn's SIGWINCH handling; we can choose another one.
 -    signal.signal(signal.SIGWINCH, make_winch_handler(server))
 -
+ # Install limits on virtual memory and CPU time consumption.
+ def pre_request(worker, req):
+     import os
+     import psutil
+     import resource
+     import signal
+     # VMS and RLIMIT_AS are the same thing: virtual memory, a.k.a. address space
+     rss, vms = psutil.Process(os.getpid()).get_memory_info()
+     soft, hard = resource.getrlimit(resource.RLIMIT_AS)
+     resource.setrlimit(resource.RLIMIT_AS, (config['virtual_memory_limit'], hard))
+     r = resource.getrusage(resource.RUSAGE_SELF)
+     cpu_time = r.ru_utime + r.ru_stime
+     signal.signal(signal.SIGXCPU, time_expired)
+     soft, hard = resource.getrlimit(resource.RLIMIT_CPU)
+     resource.setrlimit(resource.RLIMIT_CPU, (cpu_time + config['cpu_time_limit'], hard))
+ # Reset the worker if it consumes too much memory (e.g. caused by a memory leak).
+ def post_request(worker, req, environ):
+     import os
+     import psutil
+     rss, vms = psutil.Process(os.getpid()).get_memory_info()
+     if vms > config['virtual_memory_reset']:
+         _logger.info('Virtual memory consumption '
+             'too high, rebooting the worker.')
+         worker.alive = False # Commit suicide after the request.
 -# Our signal handler will signal a SGIQUIT to all workers.
 -def make_winch_handler(server):
 -    def handle_winch(sig, fram):
 -        server.kill_workers(signal.SIGQUIT) # This is gunicorn specific.
 -    return handle_winch
 -
+ # SIGXCPU (exceeded CPU time) signal handler will raise an exception.
+ def time_expired(n, stack):
+     _logger.info('CPU time limit exceeded.')
+     raise Exception('CPU time limit exceeded.') # TODO one of openerp.exception
 -# Kill gracefuly the workers (e.g. because we want to clear their cache).
 -# This is done by signaling a SIGWINCH to the master process, so it can be
 -# called by the workers themselves.
 -def kill_workers():
 -    try:
 -        os.kill(arbiter_pid, signal.SIGWINCH)
 -    except OSError, e:
 -        if e.errno == errno.ESRCH: # no such pid
 -            return
 -        raise
 -
 -class kill_workers_cache(openerp.tools.ormcache):
 -    def clear(self, dbname, *args, **kwargs):
 -        kill_workers()
 -
+ # vim:expandtab:smartindent:tabstop=4:softtabstop=4:shiftwidth=4: