mirror of
https://github.com/Sneed-Group/Poodletooth-iLand
synced 2024-12-25 04:32:33 -06:00
187 lines
6.7 KiB
Python
187 lines
6.7 KiB
Python
from direct.showbase.DirectObject import *
|
|
from pandac.PandaModules import *
|
|
from direct.task import Task
|
|
from direct.distributed import DistributedObject
|
|
from direct.directnotify import DirectNotifyGlobal
|
|
from direct.distributed.ClockDelta import globalClockDelta
|
|
|
|
class TimeManager(DistributedObject.DistributedObject):
|
|
"""
|
|
This DistributedObject lives on the AI and on the client side, and
|
|
serves to synchronize the time between them so they both agree, to
|
|
within a few hundred milliseconds at least, what time it is.
|
|
|
|
It uses a pull model where the client can request a
|
|
synchronization check from time to time. It also employs a
|
|
round-trip measurement to minimize the effect of latency.
|
|
"""
|
|
|
|
notify = DirectNotifyGlobal.directNotify.newCategory("TimeManager")
|
|
|
|
# The number of seconds to wait between automatic
|
|
# synchronizations. Set to 0 to disable auto sync after
|
|
# startup.
|
|
updateFreq = base.config.GetFloat('time-manager-freq', 1800)
|
|
|
|
# The minimum number of seconds to wait between two unrelated
|
|
# synchronization attempts. Increasing this number cuts down
|
|
# on frivolous synchronizations.
|
|
minWait = base.config.GetFloat('time-manager-min-wait', 10)
|
|
|
|
# The maximum number of seconds of uncertainty to tolerate in
|
|
# the clock delta without trying again.
|
|
maxUncertainty = base.config.GetFloat('time-manager-max-uncertainty', 1)
|
|
|
|
# The maximum number of attempts to try to get a low-latency
|
|
# time measurement before giving up and accepting whatever we
|
|
# get.
|
|
maxAttempts = base.config.GetInt('time-manager-max-attempts', 5)
|
|
|
|
# A simulated clock skew for debugging, in seconds.
|
|
extraSkew = base.config.GetInt('time-manager-extra-skew', 0)
|
|
|
|
if extraSkew != 0:
|
|
notify.info("Simulating clock skew of %0.3f s" % extraSkew)
|
|
|
|
reportFrameRateInterval = base.config.GetDouble('report-frame-rate-interval', 300.0)
|
|
|
|
def __init__(self, cr):
|
|
DistributedObject.DistributedObject.__init__(self, cr)
|
|
|
|
self.thisContext = -1
|
|
self.nextContext = 0
|
|
self.attemptCount = 0
|
|
self.start = 0
|
|
self.lastAttempt = -self.minWait*2
|
|
|
|
### DistributedObject methods ###
|
|
|
|
def generate(self):
|
|
"""
|
|
This method is called when the DistributedObject is reintroduced
|
|
to the world, either for the first time or from the cache.
|
|
"""
|
|
DistributedObject.DistributedObject.generate(self)
|
|
|
|
self.accept('clock_error', self.handleClockError)
|
|
|
|
if self.updateFreq > 0:
|
|
self.startTask()
|
|
|
|
def announceGenerate(self):
|
|
DistributedObject.DistributedObject.announceGenerate(self)
|
|
self.cr.timeManager = self
|
|
self.synchronize("TimeManager.announceGenerate")
|
|
|
|
def disable(self):
|
|
"""
|
|
This method is called when the DistributedObject is removed from
|
|
active duty and stored in a cache.
|
|
"""
|
|
self.ignore('clock_error')
|
|
self.stopTask()
|
|
taskMgr.remove('frameRateMonitor')
|
|
if self.cr.timeManager is self:
|
|
self.cr.timeManager = None
|
|
DistributedObject.DistributedObject.disable(self)
|
|
|
|
def delete(self):
|
|
"""
|
|
This method is called when the DistributedObject is permanently
|
|
removed from the world and deleted from the cache.
|
|
"""
|
|
DistributedObject.DistributedObject.delete(self)
|
|
|
|
### Task management methods ###
|
|
|
|
def startTask(self):
|
|
self.stopTask()
|
|
taskMgr.doMethodLater(self.updateFreq, self.doUpdate, "timeMgrTask")
|
|
|
|
def stopTask(self):
|
|
taskMgr.remove("timeMgrTask")
|
|
|
|
def doUpdate(self, task):
|
|
self.synchronize("timer")
|
|
# Spawn the next one
|
|
taskMgr.doMethodLater(self.updateFreq, self.doUpdate, "timeMgrTask")
|
|
return Task.done
|
|
|
|
### Automatic clock error handling ###
|
|
|
|
def handleClockError(self):
|
|
self.synchronize("clock error")
|
|
|
|
### Synchronization methods ###
|
|
|
|
def synchronize(self, description):
|
|
"""synchronize(self, string description)
|
|
|
|
Call this function from time to time to synchronize watches
|
|
with the server. This initiates a round-trip transaction;
|
|
when the transaction completes, the time will be synced.
|
|
|
|
The description is the string that will be written to the log
|
|
file regarding the reason for this synchronization attempt.
|
|
|
|
The return value is true if the attempt is made, or false if
|
|
it is too soon since the last attempt.
|
|
"""
|
|
now = globalClock.getRealTime()
|
|
|
|
if now - self.lastAttempt < self.minWait:
|
|
self.notify.debug("Not resyncing (too soon): %s" % (description))
|
|
return 0
|
|
|
|
self.talkResult = 0
|
|
self.thisContext = self.nextContext
|
|
self.attemptCount = 0
|
|
self.nextContext = (self.nextContext + 1) & 255
|
|
self.notify.info("Clock sync: %s" % (description))
|
|
self.start = now
|
|
self.lastAttempt = now
|
|
self.sendUpdate("requestServerTime", [self.thisContext])
|
|
|
|
return 1
|
|
|
|
|
|
def serverTime(self, context, timestamp):
|
|
"""serverTime(self, int8 context, int32 timestamp)
|
|
|
|
This message is sent from the AI to the client in response to
|
|
a previous requestServerTime. It contains the time as
|
|
observed by the AI.
|
|
|
|
The client should use this, in conjunction with the time
|
|
measurement taken before calling requestServerTime (above), to
|
|
determine the clock delta between the AI and the client
|
|
machines.
|
|
"""
|
|
end = globalClock.getRealTime()
|
|
|
|
if context != self.thisContext:
|
|
self.notify.info("Ignoring TimeManager response for old context %d" % (context))
|
|
return
|
|
|
|
elapsed = end - self.start
|
|
self.attemptCount += 1
|
|
self.notify.info("Clock sync roundtrip took %0.3f ms" % (elapsed * 1000.0))
|
|
|
|
average = (self.start + end) / 2.0 - self.extraSkew
|
|
uncertainty = (end - self.start) / 2.0 + abs(self.extraSkew)
|
|
|
|
globalClockDelta.resynchronize(average, timestamp, uncertainty)
|
|
|
|
self.notify.info("Local clock uncertainty +/- %.3f s" % (globalClockDelta.getUncertainty()))
|
|
|
|
if globalClockDelta.getUncertainty() > self.maxUncertainty:
|
|
if self.attemptCount < self.maxAttempts:
|
|
self.notify.info("Uncertainty is too high, trying again.")
|
|
self.start = globalClock.getRealTime()
|
|
self.sendUpdate("requestServerTime", [self.thisContext])
|
|
return
|
|
self.notify.info("Giving up on uncertainty requirement.")
|
|
|
|
messenger.send("gotTimeSync", taskChain = 'default')
|
|
messenger.send(self.cr.uniqueName("gotTimeSync"), taskChain = 'default')
|
|
|