Good can be better then None. We'd rather have the best guess at the correct time rather than no time at all, so rather than dropping a persisted TimeFreeze on reboot, assume the current device time is consistent with the previous device time.

By binding sockaddr into a sockaddr_storage we are actually assigning random memory bytes into the sockaddr_storage structre since sockaddr_t is actually smaller than sockaddr_storage.

Time since last NTP sync can be a good indicator of the "fresh"-ness of a vended NTP time.

Does not replace existing APIs; callers can continue to get the timestamp as a TimeInterval directly rather than a struct that wraps it.

hi everyone 👋

we are using Kronos as a dependency in Datadog iOS SDK and we were wondering if you had any plans for distributing Kronos as binary / xcframework?

Apple Silicon machines have arm64 simulators Previous configuration wasn't building for these machines Now we let the compiler decide on VALID_ARCHS by default

If specifying VALID_ARCHS in xcconfig was done out of neccessity somehow, we can add arm64 to them instead of removing.

@Reflejo

How does this behave when a network is not present, is slow or fails to complete request or user restarts app with no network.

Does clock.now go out of sync and become unreliable?

I am getting crash reports from NTPPacket, not a lot since there are thousand of users, but 10 in the last two weeks. Sometimes the user has been running the app for an hour without problems. It is also weird that the log points to line 20, which is empty (but that could be an error in the reporting).

MAIN THREAD - CRASHED CoreFoundation CFAssertMismatchedTypeID CoreFoundation CFRunLoopSourceInvalidate CFNetwork CFNetServiceBrowserStopSearch CoreFoundation CFRUNLOOP_IS_CALLING_OUT_TO_A_SOURCE0_PERFORM_FUNCTION CoreFoundation __CFRunLoopDoSource0 CoreFoundation __CFRunLoopDoSources0 SprintTimer main NTPPacket.swift:20 libdyld.dylib start

It looks like Clock.sync() does not work if I have WiFi off and cellular data (LTE) on. In particular, the "first:" and "completion:" blocks never get called. It works if WiFi is on. Am I missing something? It seems rather odd.

If the device clock changes after reboot before Kronos is updated, the "trusted" time will be wrong since we'll apply the offset to the wrong ground time. This breaks the "trusted" guarantee.

Hello,

if the lib is inclued in an app running on iOS 14 an authorisation is asked at launch in a system popup :

"NAME_OF_YOUR_APP" would like to find and connect to devices on your local network. This app will be able to discover and connect to devices on the networks you use. Don't allow - OK

This does not appear on Wi-Fi, it is displayed only if the phone is on cellular network like 4G.

This is triggered by Clock.sync()

Why Kronos wants to use local network ? How to avoid this popup in any case ?

Thanks

Origin issue: #81

Preparation step for binary distribution.

If we build Kronos with BUILD_LIBRARY_FOR_DISTRIBUTION = YES and add the output xcframework file as an asset to releases then our users can use this binary in most of the cases (old Xcode versions will still need to fallback to building from source but that's okay i guess). Does this make sense? Do you think we need to make other changes as well?

Also, you will need to add this to your release process; is this okay with you? (I'd be happy to help but this step requires admin rights i guess)

Hello everyone,

I would like to build an app that records frames and depth images throughout an AR session. For this purpose I need to have the time synchronized as best as possible.

My question is, whether there is a better way to get the ARFrame timestamps synced than to call sync before recording and safe the offset to add it later to the recorded internal timestamps.

Thanks already for your help!

In our tests with Austrian time servers we often see that using IPv6 addresses results in much slower query times (consistently 10 seconds instead of ~4 with only IPv4).

This PR adds a new parameter to the Clock.sync call and NTPClient to allow the preference of IPv4 addresses if both IPv4 and IPv6 addresses are resolved.

WARNING: ThreadSanitizer: data race (pid=97366)
  Read of size 8 at 0x0001164e9888 by thread T4:
    #0 static Clock.annotatedNow.getter <null>:2 (Kronos:x86_64+0x26a76ca)
    #1 static Clock.now.getter <null>:2 (Kronos:x86_64+0x26a747d)

  Previous write of size 8 at 0x0001164e9888 by main thread:
    #0 static Clock.stableTime.setter <null>:2 (Kronos:x86_64+0x26a6dcf)
    #1 closure #1 in static Clock.sync(from:samples:first:completion:) <null>:2 (Kronos:x86_64+0x26a827e)
    #2 partial apply for closure #1 in static Clock.sync(from:samples:first:completion:) <null>:2 (Kronos:x86_64+0x26a86b5)
    #3 $defer #1 () in closure #1 in closure #1 in NTPClient.query(pool:version:port:numberOfSamples:maximumServers:timeout:progress:) <null>:2 (Kronos:x86_64+0x26b7cf7)
    #4 closure #1 in closure #1 in NTPClient.query(pool:version:port:numberOfSamples:maximumServers:timeout:progress:) <null>:2 (Kronos:x86_64+0x26b783c)
    #5 partial apply for closure #1 in closure #1 in NTPClient.query(pool:version:port:numberOfSamples:maximumServers:timeout:progress:) <null>:2 (Kronos:x86_64+0x26bd6f8)
    #6 closure #1 in NTPClient.query(ip:port:version:timeout:numberOfSamples:completion:) <null>:2 (Kronos:x86_64+0x26b9af7)
    #7 partial apply for closure #1 in NTPClient.query(ip:port:version:timeout:numberOfSamples:completion:) <null>:2 (Kronos:x86_64+0x26bdb8e)
    #8 thunk for @escaping @callee_guaranteed (@guaranteed Data?, @unowned Double) -> () <null>:2 (Kronos:x86_64+0x26ba39d)
    #9 closure #1 in NTPClient.sendAsyncUDPQuery(to:port:timeout:completion:) <null>:2 (Kronos:x86_64+0x26bd0dd)
    #10 @objc closure #1 in NTPClient.sendAsyncUDPQuery(to:port:timeout:completion:) <null>:2 (Kronos:x86_64+0x26bd1f3)
    #11 __CFSocketPerformV0 <null>:2 (CoreFoundation:x86_64+0x92fd9)
    #12 start <null>:2 (libdyld.dylib:x86_64+0x1408)

  Location is global 'static Clock.stableTime' at 0x0001164e9888 (Kronos+0x000005b96888)

  Thread T4 (tid=1302869, running) is a GCD worker thread