read parameters from hardware before starting up server

+ send cached values on activate instead of reading from hardware

Change-Id: Ied9b29969cef3584a788fcd51d5b31946b0a3f25
Reviewed-on: https://forge.frm2.tum.de/review/18234
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>

secop/modules.py

@@ -177,11 +177,19 @@ class Module(object):
     def init(self):
         # may be overriden in derived classes to init stuff
         self.log.debug('empty init()')
-        mkthread(self.late_init)
 
-    def late_init(self):
-        # this runs async somewhen after init
-        self.log.debug('late init()')
+    def postinit(self):
+        self.log.debug('empty postinit()')
+
+    def late_init(self, started_callback):
+        '''runs after postinit of all modules
+
+        started_callback to be called when thread spawned by late_init
+        or, if not implmemented, immediately
+        '''
+
+        self.log.debug('empty late init()')
+        started_callback(self)
 
 
 class Readable(Module):
@@ -217,26 +225,33 @@ class Readable(Module):
 
     def init(self):
         Module.init(self)
-        self._pollthread = mkthread(self.__pollThread)
 
-    def __pollThread(self):
+    def late_init(self, started_callback):
+        '''start polling thread'''
+        mkthread(self.__pollThread, started_callback)
+
+    def __pollThread(self, started_callback):
         try:
-            self.__pollThread_inner()
+            self.__pollThread_inner(started_callback)
         except Exception as e:
             self.log.exception(e)
+            self.status = (self.Status.ERROR, 'polling thread could not start')
+            started_callback(self)
             print(formatExtendedStack())
 
-    def __pollThread_inner(self):
+    def __pollThread_inner(self, started_callback):
         """super simple and super stupid per-module polling thread"""
         i = 0
-        while True:
         fastpoll = self.poll(i)
+        started_callback(self)
+        while True:
             i += 1
             try:
                 time.sleep(self.pollinterval * (0.1 if fastpoll else 1))
             except TypeError:
                 time.sleep(min(self.pollinterval)
                            if fastpoll else max(self.pollinterval))
+            fastpoll = self.poll(i)
 
     def poll(self, nr=0):
         # Just poll all parameters regularly where polling is enabled

secop/protocol/dispatcher.py

@@ -390,7 +390,7 @@ class Dispatcher(object):
             self._active_connections.add(conn)
             modules = self._modules
 
-        # for initial update poll all values...
+        # send updates for all values. The first poll already happend before the server is active
         for modulename in modules:
             moduleobj = self._modules.get(modulename, None)
             if moduleobj is None:
@@ -399,22 +399,10 @@ class Dispatcher(object):
             for pname, pobj in moduleobj.parameters.items():
                 if not pobj.export:  # XXX: handle export_as cases!
                     continue
-                # WARNING: THIS READS ALL parameters FROM HW!
-                # XXX: should we send the cached values instead? (pbj.value)
-                # also: ignore errors here.
-                try:
-                    res = self._getParameterValue(modulename, pname)
-                    if res[0] == Ellipsis:  # means we do not have a value at all so skip this
-                        self.log.error(
-                                u'activate: got no value for %s:%s!' %
-                                (modulename, pname))
-                    #else:
-                        #rm = Message(EVENTREPLY, u'%s:%s' % (modulename, pname))
-                        #rm.set_result(*res)
-                        #self.broadcast_event(rm)
-                except SECOPError as e:
-                    self.log.error(u'decide what to do here! (ignore error and skip update)')
-                    self.log.exception(e)
+                # can not use announce_update here, as this will send to all clients
+                updmsg = Message(EVENTREPLY, module=moduleobj.name, parameter=pname)
+                updmsg.set_result(pobj.export_value(), dict(t=pobj.timestamp))
+                conn.queue_async_reply(updmsg)
         msg.mkreply()
         conn.queue_async_reply(msg)  # should be sent AFTER all the ^^initial updates
         return None

secop/server.py

@@ -33,6 +33,11 @@ try:
 except ImportError:
     import ConfigParser as configparser  # py2
 
+try:
+    from queue import Queue # py 3
+except ImportError:
+    from Queue import Queue # py 2
+
 from daemon import DaemonContext
 
 try:
@@ -202,11 +207,23 @@ class Server(object):
             self._dispatcher.register_module(devobj, devname, export)
             # also call init on the modules
             devobj.init()
-        # call a possibly empty postinit on each module after registering all
+        # call postinit on each module after registering all
         for _devname, devobj, _export in devs:
-            postinit = getattr(devobj, u'postinit', None)
-            if postinit:
-                postinit()
+            devobj.postinit()
+        starting_modules = set()
+        finished_modules = Queue()
+        for _devname, devobj, _export in devs:
+            starting_modules.add(devobj)
+            devobj.late_init(started_callback=finished_modules.put)
+        # remark: it is the module implementors responsibility to call started_callback
+        # within reasonable time (using timeouts). If we find later, that this is not
+        # enough, we might insert checking for a timeout here, and somehow set the remaining
+        # starting_modules to an error state.
+        while starting_modules:
+            finished = finished_modules.get()
+            self.log.info(u'%s has started' % finished.name)
+            # use discard instead of remove here, catching the case when started_callback is called twice
+            starting_modules.discard(finished)
 
     def _processInterfaceOptions(self, interfaceopts):
         # eval interfaces