I'm using GPUImageFilter in a chain, and most of the time it works OK. I've recently come across a few random crashes that match the symptoms in this github issue (albeit I'm using GPUImageFilter not live capture or video). I'm trying to find a suitable method that can ensure I've cleared the frame buffer and any other GPUImage-related activities in willResignActive.
Currently I have:
[[GPUImageContext sharedFramebufferCache] purgeAllUnassignedFramebuffers];
Is this sufficient? Should I use something else instead/in addition to?
As indicated there, seeing gpus_ReturnNotPermittedKillClient in a stack trace almost always is due to OpenGL ES operations being performed while your application is in the background or is just about to go to the background.
To deal with this, you need to guarantee that all GPUImage-related work is finished before your application heads to the background. You'll want to listen for delegate notifications that your application is heading to the background, and make sure all processing is complete before that delegate callback exits. The suggestion there by henryl is one way to ensure this. Add the following near the end of your delegate callback:
runSynchronouslyOnVideoProcessingQueue(^{
// Do some operation
});
What that will do is inject a synchronous block into the video processing pipeline (which runs on a background queue). Your delegate callback will block the main thread at that point until this block has a chance to execute, guaranteeing that all processing blocks before it have finished. That will make sure all pending operations are done (assuming you don't add new ones) before your application heads to the background.
There is a slight chance of this introducing a deadlock in your application, but I don't think any of my code in the processing pipeline calls back into the main queue. You might want to watch out for that, because if I do still have something in there that does that, this will lock your application. That internal code would need to be fixed if so.
Related
I'm currently trying to use Rust C interop in order to send actional desktop notifications (using a modified version of this lib). The main goal here would be to handle these notification sending events on separate threads like so:
thread::spawn(move || unsafe {
sys::sendNotification(
NSString::from_str(&title).deref(),
NSString::from_str(&message).deref(),
NSString::from_str(&uri).deref(),
NSString::from_str(&img.unwrap_or_default()).deref(),
);
});
This would allow me to have multiple notification 'handlers' running at the same time (vs. just being able to have a single notification displayed at once), and would also allow my main process to run without being blocked. Given the nature of the program (web-scraper), I don't want scraping halted whenever a notification is being displayed.
That said, this approach is somewhat problematic because the underlying obj-c code relies on NSRunLoop to handle click events (e.g., user clicks on the action to open a web page) through the created NotificationCenterDelegate instance. Per my knowledge (feel free to fact-check me on this I'm not familiar with obj-c), NSRunLoops only operate on the main thread and this code is rendered useless if ran on a worker... The notification still sends in this scenario, but events aren't processed.
Is there a way to handle this that is more effective than running my scraping logic on a separate loop and sending notif-send events to the main thread for processing (which will probably be halted by a notification that I hadn't opened)?
Strictly speaking, there is (or can be) one NSRunLoop per thread, not only the main thread. But it's still the case that GUI stuff generally needs to run on the main thread.
I recommend that you take the approach of running scraping on a separate thread. This is generally a good idea for any combination of long-running work and GUI — it ensures that the work cannot cause the UI to hang or hiccup.
When my application loads, using the didFinishLaunchingWithOptionsi parse data from the internet to nsarrays. My question is, when the user exists the application by using the 'home' button, and then loads the application again how can the data be re-loaded? (because if the data does not reload - if there are any updates on websites, the new updates will not be seen).
Add an applicationWillEnterForeground method to your app delegate. Load the data there, or start a thread to load it if you like.
You should probably also periodically check for new data even while the app remains open, because the user could go idle for a long time.
As an aside, you shouldn't do anything which might block in applicationDidFinishLaunchingWithOptions. If you are using synchronous NSURLConnection APIs there is a danger the OS might kill your app for taking too long to launch. Best to either use the asynchronous/NSURLConnectionDelegate APIs or do the networking on a background thread and call back to the main thread when you need to update UI (UIKit does NOT like being called from background threads, as it is not thread safe. It might appear to work sometimes, but it will come back to bite you sooner or later).
In an iOS app, I'm writing a class that will be messaged, go do a background request (via performSelectorInBackground:withObject:), and then return the result through a delegate method (that will then be displayed on a map). Everything seems to work right when one request is happening at a time, but I'm trying to figure out how to handle multiple overlapping requests. For example, if a user enters something in a search box that starts a background thread, and then enters something else before the initial background thread completes, how should this be handled?
There are a few options (don't let the second request start while the first is in progress, stop the first as soon as the second is requested, let both run simultaneously and return independent results, etc.), but is there a common/recommended way to deal with this?
I don't think there's universal answer to this. My suggestion is to separate tasks (in form of NSOperations and/or blocks) by their function and relationships between them.
Example: you don't want add image resizing operation to the same queue with fetching some unrelated feed from web, especially if no relationship between them exists. Or maybe you do because both require great amount of memory and because of that can't run in parallel.
But you'd probably want to add web image search operations to same queue while canceling operations of the same type added to this queue before. Each of those image search operations might init image resize operation and place it into some other queue. Now you have an relationship and have to cancel resizing in addition to image search operation. What if image search operation takes longer than associated resize operation? How do you keep a reference to it or know when it's done?
Yeah, it gets complicated easily and sorry if I didn't give you any concrete answers because of uniqueness of each situation but making it run like a Swiss clock in the end is very satisfying :)
Using the iPhone and objective C, is there a way to stall or perform a timing loop to allow for the GPS to catch up and return a valid set of coordinates?
Currently, the application runs too quickly and the GPS cannot supply the coordinates fast enough...
Since you said you're on iPhone, you're using CLLocationManager. Just set a delegate on the manager and wait for the locationManager:didUpdateToLocation:fromLocation: message to know when the GPS data is ready.
Assuming your GPS polling is running in a different thread to the User Interface, you can call the static NSThread functions sleepForTimeInterval or sleepUntilDate from the thread that is waiting for the GPS data.
If your mobile application is using GPS, your application should be prepared for location updates, even if your application doesn't track movements..
A common case would be where the user put your application in background and activate it later on a completely different location.
On iOS, create an implementation of CLLocationManagerDelegate like Anomie wrote. And use the timestamp of the update to evaluate the freshness of the location.
Don't sleep & poll like other people suggested.
Either block to wait for data or don't update anything if no data received. There is of course usleep(), but without showing code and specifically how your loop is executed and by what mechanism (threaded or not) we can only answer in general terms.
Scenario:
I have a Distributed-objects-based IPC between a mac application and a launchd daemon (written with Foundation classes). Since I had issues before regarding asynchronous messaging (e.g. I have a registerClient: on the server's root object and whenever there's an event the server's root object notifies / calls a method in the client's proxy object), I did long-polling which meant that the client "harvests" lists of events / notifications from the daemon. This "harvest" is done through a server object method call, which then returns an NSArray instance.
It works pretty well, until for a few seconds, the server object's process (launched thru launchd) starts being labeled red with the "(Not responding)" tag beside it (inside Activity Monitor). Like I said, functionally, it works well, but we just want to get rid of this "Not responding" label.
How can I prevent this "Not responding" tag?
FYI, I already did launchd-based processes before and this is the first time I did long-polling. Also, I tried NSSocketPortNameServer-based connections and also NSSocketPort-based ones. They didn't have this problem. Locking wasn't also an issue 'coz the locks used were only NSCondition's and we logged and debugged the program and it seems like the only locking "issue" is on the harvesting part, which actually, functionally works. Also, client-process is written in PyObjC while server process was written using ObjC.
Thanks in advance.
Sample the process to find out what it's doing or waiting on.
Peter's correct in the approach, though you may be able to figure it out through simple inspection. "Not responding" means that you're not processing events on your event queue for at least 5 seconds (used to be 2 seconds, but they upped it in 10.4). For a UI process, this would create a spinning wait cursor, but for a non-UI process, you're not seeing the effects as easily.
If this is a runloop-based program, it means you're probably doing something with a blocking (synchronous) operation that should be done with the run loop and a callback (async). Alternately, you need a second thread to process your blocking operations so your mainthread can continue to respond to events.
My problem was actually the call for getting a process's PID using the signature FNDR... that part caused the "Not responding" error and it never was the locks or the long-polling part. Sorry about this guys. But thank God I already found the answer.