[server.py] rename initialisation steps for better clarity

initialisation occurs in this order:
 - object creeation (via __init__ which should consume the cfg values it knows about)
 - registering each object with the dispatcher
 - calling init_module() on each module (for connecting to other modules, checking hw, creating threads....)
 - calling start_module(cb) on each module. after the module finished startup it should call cb(self) once.
   This is the right place to do initialisation of hw which is not needed to read from the hw.
   (uploading curves, polling/re-setting all parameters, etc.)

Change-Id: Ieaf9df5876e764634836861241f58ab986027f44
Reviewed-on: https://forge.frm2.tum.de/review/18566
Tested-by: JenkinsCodeReview <bjoern_pedersen@frm2.tum.de>
Reviewed-by: Markus Zolliker <markus.zolliker@psi.ch>
Reviewed-by: Enrico Faulhaber <enrico.faulhaber@frm2.tum.de>
This commit is contained in:
Enrico Faulhaber
2018-07-31 16:37:36 +02:00
parent acb8e6d0a1
commit 5f640ce299
9 changed files with 52 additions and 41 deletions

View File

@ -214,16 +214,16 @@ class Server(object):
for devname, devobj, export in devs:
self.log.info(u'registering module %r' % devname)
self._dispatcher.register_module(devobj, devname, export)
# also call init on the modules
devobj.init()
# call postinit on each module after registering all
# also call early_init on the modules
devobj.early_init()
# call init on each module after registering all
for _devname, devobj, _export in devs:
devobj.postinit()
devobj.init_module()
starting_modules = set()
finished_modules = Queue()
for _devname, devobj, _export in devs:
starting_modules.add(devobj)
devobj.late_init(started_callback=finished_modules.put)
devobj.start_module(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
@ -233,6 +233,7 @@ class Server(object):
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)
finished_modules.task_done()
def _processInterfaceOptions(self, interfaceopts):
# eval interfaces