Last active
March 19, 2024 14:40
-
-
Save roostr/ead83858291ce3ce214fcf8ead3cd825 to your computer and use it in GitHub Desktop.
A watchdog timer for iOS to detect when the main run loop is stalled
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
// | |
// Watchdog.swift | |
// | |
// The MIT License (MIT) | |
// Copyright © 2023 Front Pocket Software LLC | |
// | |
// Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated | |
// documentation files (the “Software”), to deal in the Software without restriction, including without limitation the | |
// rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to | |
// permit persons to whom the Software is furnished to do so, subject to the following conditions: | |
// The above copyright notice and this permission notice shall be included in all copies or substantial portions of the | |
// Software. | |
// THE SOFTWARE IS PROVIDED “AS IS”, WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE | |
// WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS | |
// OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR | |
// OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. | |
// This class will monitor your app's run loop. It detects when a runloop starts and then fails to complete | |
// within TIMEOUT seconds. If this threshhold is exceeded, it will call timeoutHandler(). | |
// | |
// You probably want to set your own timeoutHandler so that you can do something about this. Consider calling | |
// abort(), as that will send a crash report to Apple on devices that have diagnostic reporting turned on | |
// (which is typically about 20% of devices). Then, you can look at the stacktrace of the main thread to see | |
// where the main thread was around the time the TIMEOUT threshhold was exceeded. | |
// | |
// To ensure that access to currentRunLoopStartTime from concurrent threads behaves in a safe way, we use | |
// a serial DispatchQueue for all reads and writes to currentRunLoopStartTime. | |
// | |
// USAGE | |
// 1. Add this file to your project | |
// 2. If desired, tweak the TIMEOUT value to be different (default is 4 seconds) | |
// 3. Set your own timeoutHandler to override the default one | |
// 4. Call Watchdog.shared.start() | |
// 5. (optionally) call stop() when you enter background, start() when you come back to foreground | |
// (I don't know whether or not step 5 is necessary) | |
// | |
// TO VERIFY | |
// Put a `sleep(50)` into one of your button handlers, fire up your app, and tap that button. | |
// | |
// KNOWN ISSUES | |
// 1. You are not guaranteed to get your callback. We know that if the main run loop is stuck, something funky is going | |
// on. It is not known whether the serialQueue.asyncAfter task will fire, because it does not have a dedicated | |
// thread to run on, as it is sharing a thread pool. In my testing, this code worked reliably, but it's not | |
// guaranteed to work 100% of the time. | |
// 2. I haven't tested this in production yet. I will probably only add logging in my initial rollout of it to ensure | |
// it isn't hitting false-positives. So long as it's not, I plan to add the call to abort() in a future release. | |
// I'm going to call abort() so I can get a callstack of the main thread reported back in Xcode's crash reporting. | |
// Even then, I plan to only call abort() once per device (by remembering whether it was called with a UserDefaults | |
// flag), just as an extra protection against potential false positives and bad user experiences. | |
import Foundation | |
import UIKit | |
@objc final public class Watchdog: NSObject { | |
@objc public static let shared = Watchdog() | |
fileprivate static let TIMEOUT: TimeInterval = 4 | |
fileprivate static let NOT_IN_RUN_LOOP: TimeInterval = 0 | |
fileprivate let serialQueue = DispatchQueue(label: "watchdog.queue") | |
@objc public var timeoutHandler: () -> Void = { print("Listen: Main Run Loop has come unstuck in time.") } | |
fileprivate var runLoopStartTime: TimeInterval = NOT_IN_RUN_LOOP | |
fileprivate var isRunning = false | |
fileprivate lazy var runLoop: CFRunLoop = { | |
CFRunLoopGetMain() | |
}() | |
fileprivate lazy var runLoopObserver: CFRunLoopObserver = { | |
CFRunLoopObserverCreateWithHandler(kCFAllocatorDefault, CFRunLoopActivity.allActivities.rawValue, true, .min) { _, activity in | |
switch activity { | |
case .entry, .beforeTimers, .afterWaiting, .beforeSources: | |
self.serialQueue.async { | |
if self.runLoopStartTime == Self.NOT_IN_RUN_LOOP { | |
self.runLoopStartTime = Date().timeIntervalSince1970 | |
} | |
} | |
case .beforeWaiting, .exit: | |
self.serialQueue.async { | |
self.runLoopStartTime = Self.NOT_IN_RUN_LOOP | |
} | |
default: | |
break | |
} | |
} | |
}() | |
deinit { | |
stop() | |
CFRunLoopObserverInvalidate(self.runLoopObserver) | |
} | |
@objc func start() { | |
guard !isRunning else { return } | |
isRunning = true | |
serialQueue.async { | |
self.runLoopStartTime = Self.NOT_IN_RUN_LOOP | |
} | |
CFRunLoopAddObserver(self.runLoop, self.runLoopObserver, CFRunLoopMode.commonModes) | |
enqueueBackgroundCheck() | |
} | |
@objc func stop() { | |
guard isRunning else { return } | |
isRunning = false | |
serialQueue.async { | |
self.runLoopStartTime = Self.NOT_IN_RUN_LOOP | |
} | |
CFRunLoopRemoveObserver(self.runLoop, self.runLoopObserver, CFRunLoopMode.commonModes) | |
} | |
fileprivate func enqueueBackgroundCheck(after: TimeInterval = TIMEOUT) { | |
guard isRunning else { return } | |
serialQueue.asyncAfter(deadline: .now() + after) { | |
guard self.runLoopStartTime != Self.NOT_IN_RUN_LOOP else { | |
Watchdog.shared.enqueueBackgroundCheck() | |
return | |
} | |
let now = Date().timeIntervalSince1970 | |
let elapsedTime = now - self.runLoopStartTime | |
if elapsedTime >= Self.TIMEOUT { | |
self.timeoutHandler() | |
self.stop() | |
} else { | |
// How far into the current run loop are we? If the current runloop has already run | |
// for 1.5 seconds, then we should check back in after "4 - 1.5 = 2.5" seconds. | |
Watchdog.shared.enqueueBackgroundCheck(after: Self.TIMEOUT - elapsedTime) | |
} | |
} | |
} | |
} |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Updated to add this to start:
Because it can't hurt, and it could address a possible race condition where
stop()
is called (which clears out therunLoopStartTime
, but then a final run loop observer sneaks in after that and setsrunLoopStartTime
to a valid value. And then, if we calledstart()
while the run loop wasn't executing (like the background), enqueueBackgroundCheck may think we are stuck because it's looking at a stalerunLoopStartTime
value.So to avoid all that, we just reset the
runLoopStartTime
to be a sane value every time we callstart()
.