new iOS guy here. I have a problem that Googling and searching on here has not shed any light on. I'm assuming this is basic. I have a simple app (app delegate and 1 view controller), and as part of it I'm using local notification. So, in the app delegate I use the 'didReceiveLocalNotification' to watch for the notifications. Depending on which one comes in, I then call one of several instance methods in my main view controller.
ie in the AppDelegate.m
- (void)application:(UIApplication *)application didReceiveLocalNotification: (UILocalNotification *)notification {
MyViewController* controller = [[MyViewController alloc] autorelease];
if ([[notification.userInfo objectForKey:#"id"] isEqualToString:#"myKey"]) {
[controller checkActive];
}
}
Through logging and watching some breakpoints, this is all working. If the app is in the background, the notification comes in, app opens, and the correct instance method is called.
What I cannot figure out at all is why some parts of the instance method are simply being passed by, with no effect. For a simple example, if we have this:
-(void)checkActive {
ViewThing.alpha = 1.0;
NSLog(#"checkActive ran");
}
The log statement will show up fine, but the ViewThing will not change. Elsewhere in the main view controller I'm calling the same checkActive method with no problems and it changes the ViewThing. (via another interface button IBAction method in that case).
There are no errors, no warnings, the console shows nothing, putting a breakpoint on ViewThing shows that it hits the line. I'm stumped, cannot see what is different from trying to calling the method from the delegate vs. on an IBAction.
Thanks for any tips!
If the alpha is not correctly changing there a few possible issues with 1 and 2 being the most likely.
ViewThing is nil. Reasons could be is the view unloaded and you set it to nil or checkActive was called before the outlets were set.
ViewThing.alpha is being set on a thread that is not the main thread. Attempting to change UI elements on a separate thread will caused undefined behavior such as never updating the change or taking an extended amount of time to update the change. You can check if it is the main thread using [NSThread isMainThread].
ViewThing is pointing a different view.
1 & 2 can easily be checked by logging view
NSLog(#"checkActive ran %#", ViewThing);
Related
I'm running into a problem where 9 times out of ten, when I call UIAlertView's dismissWithClickedButtonIndex:animated:, the delegate method alertView:willDismissWithButtonIndex: is not called. Is anyone else running into this problem? I'm about to file a bug with Apple but I'm curious to see if anyone else has run into this issue and figured out any workarounds.
To ensure a consistent behavior across iOS4 and 5, you could just remove the UIAlertView's delegate just prior to calling its dismissWithClickedButtonIndex:animated: method, then manually invoke the delegate method. e.g.
- (void)somethingDidHappen {
id<UIAlertViewDelegate> delegate = myAlertView.delegate;
myAlertView.delegate = nil;
// now, we know the delegate won't be called...
[myAlertView dismissWithClickedButtonIndex:0 animated:NO];
// ...so we call it ourselves below
[delegate alertView:myAlertView clickedButtonAtIndex:0];
}
(That code isn't tested, but you get the point.)
Delegates of UI objects are only called when the user performs an action. Apple assumes that when you do something from code, you already know what you're doing and you don't need to be informed. That applies to all delegates (scrolling delegate methods of UIScrollView vs. code-scrolling, Table View manipulation, ...)
Anyway, what button index should the delegate be called with?.. there is no one when you dismiss programmatically
According to Why doesn't dismissWithClickedButtonIndex ever call clickedButtonAtIndex? the problem is that a different method is being called. However, that doesn't explain why you get erratic calls. On the devices I tested the dismiss method gets called correctly, so I only redirect it to the click version.
Maybe you should file a bug with Apple if you continue seeing the erratic behaviour.
There are alertView:clickedButtonAtIndex:, alertView:didDismissWithButtonIndex: and alertView:willDismissWithButtonIndex:. The method that you're referring to (clickedButtonAtIndex:) is only called when the user explicitly taps on a button on your alert view (hence 'clicked').
Programmatic calls via dismissWithClickedButtonIndex:animated: to dismiss the alert does not seem to call alertView:clickedButtonAtIndex:.
So, if you need some behavior to be always triggered upon the dismissal of the alert view—whether it was triggered by the user tapping on a button or triggered programmatically—then using the didDismissWithButtonIndex: and willDismissWithButtonIndex: makes more sense.
I have a navigation based app. Press a button on main view, then I push a new view to the navigation controller. All pretty basic stuff.
When the new view is loaded, I do an ASIHTTPRequest to fetch some json data, which is a list of image urls.
Then I do a for loop, create a bunch of ASIHTTPRequests, add them to a queue and then run the queue.
But if I click on the back button before the queue is finished, the app crashes, this app displays houses and lets say you pick the wrong house, click back very quickly, before any photo is displayed, bumm crash.
This thread http://groups.google.com/group/asihttprequest/browse_thread/thread/3d4815198aa889b9 explains my problem real well, except I do cancel all requests on view did unload, set delegate to nil and release the queue.
Still I crash. I crash pretty much every time if I use 3G, but on wifi it is real hard to make it crash, but quite doable.
In almost 80% instances the debugger jumps to this line in ASIHTTPRequest.m
(void)requestReceivedResponseHeaders:(NSMutableDictionary *)newResponseHeaders {
if ([self error] || [self mainRequest]) { return; }
--> if (delegate && [delegate respondsToSelector:didReceiveResponseHeadersSelector]) {
Many many cases it jumps to :
(void)requestReceivedResponseHeaders:(NSMutableDictionary *)newResponseHeaders {
if ([self error] || [self mainRequest]) { return; }
---> if (delegate && [delegate respondsToSelector:didReceiveResponseHeadersSelector]) {
Ad in a handful of instances it goes to my main loop
int main(int argc, char *argv[]) {
NSAutoreleasePool * pool = [[NSAutoreleasePool alloc] init];
--> int retVal = UIApplicationMain(argc, argv, nil, nil); with SIGBART error [pool release]; return retVal;
I am using MBP and MacPro, latest OS X, Xcode 4.0.2 and I test on all apple devices except original iPhones.
I really don't want to re-write my whole app, but is there anything else out there that compares with ASIHTTPRequest ?
Try cancelling and unsetting the delegate in -viewWillUnload rather than -viewDidUnload. I suspect the window of time in which it's actually unloading (between calling those two UIViewController methods) is the time period when you're crashable. The delegate has gone away, but you haven't told your ASIHTTPRequest object that yet.
The error is that the delegate is still set.
I have found 2 ways to fix this.
The way I consider ugly is that you make a universal delegate that does all network traffic and is instantiated when the app is first run. I actually used the app delegate and listen to nsnotification center messages. It works like a charm, the app never crashes, but I think it is not optimal.
The best way is to not set the delegate and not use "setDidFinishSelector", but instead use "setCompletionBlock:^". This will only work on devices running iOS 4.0 and up, which is more than 90-95% and growing. This is just an awesome way and will not crash the application.
You won't find anything better that ASIHTTPRequest, the problem will be how you are using it and vanishing delegates on navigation are a common problem to have to deal with.
It sounds like your problem relates to the viewcontroller that is handling the queue being destroyed due to user navigation. I find the best way of solving these issues is to have a central model class that handles all my communications and keep that class throughout the application lifecycle.
That way you don't get unexplained crashes when delegates have vanished unexpectedly.
Option 2
Another approach can be to disable user navigation until the network operation completes. Put a modal view over the entire screen that shows a uiactivityview so the user knows their actions are being blocked. Then you can fade the modal view off when the data has arrived. If you design the screen nicely with a gradient so the background just dims a bit, this can look OK. But it's not really the best approach - you should fix the delegate AWOL instead.
We probably need to see more of the code relating to the queue creation, destruction etc to find the exact issue.
Your application delegate can own an array of request queues. The array lives independent of the state of the navigation controller stack and associated views. Instead of tying requests to a view controller in the nav stack, and having to do UI tricks to block popping back to a parent view, you could add requests to an app delegate queue instance, or stop all requests and empty the queue, etc.
In my app, I have a UITableViewController which loads calculated data from my Core Data store. This can take quite some time and hangs the main thread, so to give the user visual feedback, I have installed the MBProgressHUD widget which will show a progress indicator (just the spinning wheel) while the calculation method runs in a separate thread.
This is fine, except that the user can still exit the UITableViewController if they think it's taking too long. Of course, this is a good thing, except that when the separate thread concludes its operation, it still tries to call its delegate (the UITableViewController) to hide the MBProgressHUD. This causes a crash because since the user already exited the UITableViewController, it has dealloc'd and release'd itself.
MBProgressHUD has the following code to try to stop this:
if(delegate != nil && [delegate conformsToProtocol:#protocol(MBProgressHUDDelegate)]) {
if([delegate respondsToSelector:#selector(hudWasHidden)]) {
[delegate performSelector:#selector(hudWasHidden)];
}
}
However, my app somehow seems to still be running this inner code ([delegate performSelector:#selector(hudWasHidden)]) even though the UITableViewController is totally gone--causing the app to crash.
Any suggestions? I am not running with NSZombiesEnabled.
From your UITableViewController viewWillDisappear, viewDidDisappear or dealloc method, set the MBProgressHUD.delegate = nil;
Once the user has exited the table controller, the delegate property of the hud points to a deallocated object (= a memory zone that could contain anything). That causes the crash when the computing thread ends and tries to send any message to the delegate.
In your table view controller dealloc, you must set the hud delegate to nil.
I have a navigation controller. One of the views adds custom subviews in its viewDidAppear:. I notice that the first time I navigate to an instance of this view controller after launching the app, viewDidAppear: invokes twice. If I pop this view off the stack and navigate to it again, viewDidAppear: invokes only once per appearance. All subsequent appearances invoke viewDidAppear: once.
The problem for me is that the first time I get to this view I end up with twice the number of subviews. I work around this problem by introducing a flag variable or some such, but I'd like to understand what is happening and how come I get two invocations in these circumstances.
You should never rely on -viewWillAppear:/-viewDidAppear: being called appropriately balanced with the disappear variants. While the system view controllers will do the best they can to always bracket the calls properly, I don't know if they ever guarantee it, and certainly when using custom view controllers you can find situations where these can be called multiple times.
In short, your -viewWillAppear:/-viewDidAppear: methods should be idempotent, meaning if -viewDidAppear: is called twice in a row on your controller, it should behave properly. If you want to load custom views, you may want to do that in -viewDidLoad instead and then simply put the on-screen (if they aren't already) in -viewDidAppear:.
You could also put a breakpoint in your -viewDidAppear: method to see why it's being called twice the first time it shows up.
maybe you invoke viewDidAppear in viewDidLoad (or some other stuff is going on there), since it's invoked only once during loading the view from the memory. It would match, that it's invoked two times only the first time.
This was not an iOS 5 bug, but a hidden behavior of addChildViewController:. I should file a radar for lack of documentation, I think
https://github.com/defagos/CoconutKit/issues/4
If you have a line like this in your AppDelegate
window = [[UIWindow alloc] initWithFrame:[[UIScreen mainScreen] bounds]];
make sure you DON'T have a "Main nib file base name" property in your plist set to "Window.xib" or whatever your custom window nib is named. If you do, remove that row from your plist and make sure you something like
yourRootVC = [[UIViewController alloc] init];
[window setRootViewController:yourRootVC];
in your AppDelegate after instantiating your window. In most cases, you could then safely delete the Window.xib as well.
You definitely should provide more info.
Is this the root view controller?
Maybe you initiate the navigation controller with this root view controller and then push it to the navigation controller once again?
Another solution that may have been your underlying cause: Be sure you are not presenting any new view controllers from within your viewWillAppear: method.
I was calling:
[appDel.window.rootViewController presentViewController:login animated:YES completion:nil];
from within viewWillAppear and seeing my originating VC's viewDidAppear: method called twice successively with the same stack trace as you mention. And no intermediary call to viewDidDisappear:
Moving presentViewController: to the originating VC's viewDidAppear: method cleared up the double-call issue, and so now the flow is:
Original viewDidAppear: called
Call presentViewController here
Original viewDidDisappear: called
New view is presented and no longer gives me a warning about "unbalanced VC display"
Fixed with help from this answer: https://stackoverflow.com/a/13315360/1143123 while trying to resolve "Unbalanced calls to begin/end appearance transitions for ..."
it's such an annoying problem, you'd think it runs once but then I now found out about this which is causing mayhem... This applies to all 3 (ViewDidAppear, ViewDidLoad, and ViewWillAppear), I am getting this when integrating with a payment terminal; once it finish calling the API, the window is being re-loaded when it's already on-screen and all it's memory is still there (not retained).
I resolved it by doing the following to all the routines mentioned above, below is a sample to one of them:
BOOL viewDidLoadProcessed = false;
-(void)viewDidLoad:(BOOL)animated
{
if (!viewDidLoadProcessed)
{
viewDidLoadProcessed = YES;
.
.
.
... do stuff here...
.
.
}
}
Repeat the above for all the other two, this prevents it from running twice. This never occurred before Steve Jobs died !!!
Kind Regards
Heider Sati
Adding [super viewDidAppear:animated]; worked for me:
//Called twice
- (void)viewDidAppear:(BOOL)animated{
}
//Called once
- (void)viewDidAppear:(BOOL)animated{
[super viewDidAppear:animated];
}
Resolution: While trying to recreate this bug in a fresh project to submit to Apple, I discovered that it is specific to iPhone OS 2.1, and compiling for 2.2 fixes the problem. Stephen, thanks for your help; I'll be accepting your answer since it would have worked if the bug still existed or I wasn't willing to compile for 2.2.
I have an app which is radically changing its database schema in a way that requires me to transform old-style records to new-style ones in code. Since users may store a lot of data in this app, I'm trying to display a modal view controller with a progress bar while it ports the data over (i.e. as the very first thing the user sees). This view controller's viewDidAppear: begins a database transaction and then starts a background thread to do the actual porting, which occasionally uses performSelectorInMainThread:withObject:waitUntilDone: to tell the foreground thread to update the progress bar.
The problem is, viewDidAppear: is being called twice. I noticed this because that "start a transaction" step fails with a "database busy" message, but setting a breakpoint reveals that it is indeed called two times—once by -[UIViewController viewDidMoveToWindow:shouldAppearOrDisappear:], and again by -[UIViewController modalPresentTransitionDidComplete]. Those names appear to be private UIViewController methods, so I'm guessing this is either a framework bug, or I'm doing something UIKit isn't expecting me to do.
Two relevant code excerpts (some irrelevant code has been summarized):
- (void)applicationDidFinishLaunching:(UIApplication *)application {
(register some default settings in NSUserDefaults)
// doing this early because trying to present a modal view controller
// before the view controller is visible seems to break it
[window addSubview:[self.navigationController view]];
// this is the method that may present the modal view
[self.databaseController loadDatabaseWithViewController:self.navigationController];
if(!self.databaseController.willUpgrade) {
[self restoreNavigationControllerState];
}
}
And from my DatabaseController class:
- (void)loadDatabaseWithViewController:(UIViewController*)viewController {
(open the new database)
(compute the path the old database would live at if it existed)
if([[NSFileManager defaultManager] fileExistsAtPath:oldDBPath]) {
(open the old database)
[viewController presentModalViewController:self animated:NO];
}
}
So, is there something I'm screwing up here, or should I file a bug report with Apple?
I saw this in my app too. I never got it entirely confirmed, but I think this is what's happening:
Load root view
Load modal view
OS sends view did appear notification for the view in step 1
The current view controller, which in this instance happens to be your DatabaseController class, picks it up
OS sends the view did appear notification for the modal view
The current view controller gets the notification. In this case it's the exact same controller as last time
In my case I just reset what happened in the first call to viewDidAppear:.
In your case two options spring to mind: a static variable to track whether you've started the upgrade already; or look at the UIView* parameter passed in before starting.