I'm making an app that has a few view controllers with UIDatePicker controls.
After having lots of memory warnings, I've done a lot of debugging on the cause. Using instruments, I was able to see that whenever a view controller containing a UIDatePicker appeared, memory usage was going up by 2-3mb. Ultimately, this leads to a memory warning. While I can do so much in here, and have seen a dramatic decrease in crashes since re-writing these functions, I'm still not comfortable with this behaviour and want to fix it.
The UIDatePicker's are part of the XIB file. I've tried taking them out of the XIB and programmatically adding them. This doesn't solve the problem. I've also tried using a (thread-safe" singleton. This limits memory usage to only one instance (so memory usage doesn't increase), but I'm trying to avoid singletons wherever possible. Should I just suck it up and use a singleton?
I'm seeing this behaviour on iOS 5 and 4.3, across all devices that run it. I'm using ARC, as the title suggests.
Instruments does not show any memory leaks.
Here is a screenshot from Instruments. Each step up in usage is when a VC containing a date picker appears. The baseline at the start of the graph is 2.3mb of memory usage.
http://i.stack.imgur.com/1S7ns.png
Cheers!
The tool you want to use here is called heapshot. It will help you narrow down the types of objects that are being incorrectly held onto. As #sudo notes, the problem is probably in your delegate rather than the date picker itself. You're probably not "leaking" in the strictest sense of the term. You're probably just holding onto memory that you didn't mean to (which looks a lot like a leak).
You should take a careful look at whether you have any circular strong references. Do you have objects that have a strong reference to their delegate for instance? That will typically lead to a retain loop (which will never be released under ARC). Heapshot can help you find these objects.
I seem to have solved my own question. As this was an inputView, and I also had a UIToolbar with a "Done" button as the inputAccessoryView.
So now I create these programatically in viewWillAppear. In viewWillDisappear, I set the UITextField's inputView and inputAccessoryView to nil, and the date picker and date picker toolbar to nil.
They are no longer in the XIB file.This seems to work really well - memory spikes up by 3mb, then goes back down the second the views disappear. Now the app is running consistently at under 10mb of live memory - it spikes up to 15mb on occasion but very rarely.
Thanks for the help sudo rm -rf and Rob - heapshot analysis helped to try and get to the bottom of it.
Related
I'm having a memory issue with one specific ViewController and memory. When I launch my app in debug mode, it launches with 40mb memory usage. (I have no idea if this is already a lot or not -- what is common?). Then, when I open this specific view, it spikes up to about 120mb. The issue is, when I pop this view (with the navigationController popViewController), the memory stays up at 120mb. And then, when I re-open it, it spikes to 200mb (a 80mb increase every time).
With other similar ViewController it spikes up to 120mb too, but when I close the view, it goes back down to more or less 40mb.
The problem is that this specific view contains quite a bit of code (about 1000 lines...) and it's impossible for me to post everything here.
What methods should I use to specifically locate the issue in xCode?
For anyone I might be able to help with this:
use the tools in xCode as recommended. There are some great tutorials online.
In my case it was an issue with an [NSTimer] which kept a strong reference to my view, so it never got released afterwards, thus stacking up memory. Make sure to stop your time when you pop a view.
I am going in and out of UIViewControllers.
Every time I am going in memory increases, but when I am getting out (pressing Back) memory is not decreasing.
I have weak #IBOutlets, vars and lets.
How can I fix it?
p.s. One of the #IBOutlet is container pointing to another UIViewController
Ignore the comments suggesting retain/release breakpoints and peppering your code with log calls; that's an unwieldy tactic that misses entire classes of problems. Abizern's suggestion to use Instruments is a lot more spot-on. Run the Leaks instrument to look for leaked instances. Browse the other memory instruments as well. Instruments is a &#&$&-ing magical debugging tool and so many developers ignore it for reasons I can't fathom. Spend a few minutes with a tutorial and you'll save months of debugging time over your career.
It sounds like you are creating a new UIViewController every time you switch between views. Try to create the segue to be an unwind segue, this will remove the current view from the stack. Otherwise it appears that you are just continually adding to the stack of views. Tell me if this works.
In developing my current app, I ran into some issues that I eventually traced back to a low memory warning. Part of the low memory is coming from my (liberal) use of UIWebViews, which are apparently consuming a lot of memory.
I didn't think this would be an issue, since a view that isn't currently visible should just vanish when the low memory warning is thrown, unfortunately it turns out that anything connected to the UITabBarController remains in the heirarchy and doesn't release. By manually releasing (and then recreating in viewWillAppear) the views, I make things work decently. But it doesn't completely solve the memory warning issue.
So what I need to do is manually release the view -- and the large amount of memory that winds up connected to it -- and then restore it. Since I don't want to build it programmatically (that's what IB is for!), I need to somehow reload it from storyboard.
Or, alternatively, I'm being an idiot and there's something really obvious to make my life easy.
After more experimentation, it turns out that on the one hand, my understanding of view life cycles was slightly flawed, and on the other my experiments were tainted by having accidentally let zombie objects on.
Views will, in fact, unload their contents in a low memory condition and reload them later -- that's part of what 'viewDidLoad' and 'viewDidUnload' are designed for! Putting init code in them, as some tutorials I read did, was a major error. That init code should go in the designated initializer -- even if that can be annoying to figure out which initializer is designated.
I just started learning the instruments tool, and I'm pretty sure what I am seeing is not good. When I look at allocations, specifically the column "Live Bytes" and "Overall Bytes", I see the number continually increases as the app runs...
My app has two view controllers. A table view, and the second view controller displays detailed information about the row they selected in the table view, downloaded from the internet.
I kept clicking a row in the table view, followed by clicking the back button in the navigation bar... and LiveBytes continued to increase.
I'm guessing this means that my objects aren't being released from memory... but please correct me if I'm wrong.
MY QUESTION IS: How do I use the data in instruments/allocations to track down this memory issue? How do I find the objects not being released from memory?
I'm looking for tips on how to use these tools to clean up any memory problems my app has.
Thanks!
XCODE 4.2.1, deploying to iOS 5.0+
EDIT: I'm looking at the #living column and seeing objects like UIScrollView continuously increase... and never decrease. When I click the back button in a navigation bar, are objects automatically released from memory? When are objects released, or do I need to do it manually? Or could I be running into an issue due to using strong pointers, causing objects to not be released?
Whenever you want to observe memory usage in a cyclic pattern, there's the wonderful Heapshot analysis in the "Allocations" instrument.
Start your app and go to a default state.
In Instruments, press the "Mark Heap" button to create the "Baseline".
Do something in your app like pushing a view controller.
Return to the default state.
Press the "Mark Heap" button again to create a heapshot.
Repeat about five times from step 3.
This will result in a list of heapshots, each showing the objects that are still alive from that cycle. If your app has no leaks there will be no objects left in the middle heapshots.
The first one or two cycles might have warmed up some caches, the last two might not have cleaned up some reused resources. That's why it's usually a good idea to make four to six heapshots.
The magic in the heapshot analysis lies in the fact that the heapshots show you the leaked objects from previous cycles and remove them automatically when the objects are released later. In contrary to the "Leaks" instrument it also finds abandoned memory, not only leaks.
Most Probably you have discarded the arm64 and are running your app with armv7 only. Add both arm64 and armv7 as architectures
I think one of the best ways to solve memory issues is to use ARC.
Edit -> Refactor -> Upgrade to Objective-C ARC.
ARC will handle the majority of memory management in your app. Especially sice your app doesn't sound too complex, it might totally eliminate your problem. You still need to watch out for retain cycles and listen to memory warnings though. If you don't want to use ARC (which you should) at least run the static analyzer. Your problem might be something simple that the static analyzer can show you how to fix.
Edit:
You mentioned scroll views- this might be your problem: Memory leak every time UIScrollView is released
The profile tool has an instrument called 'Leaks'. It is similar to the 'Allocations' instrument but it shows you the object that was not released. May be you can use the 'Leaks' tool to find what is the object that was retained and release these object.
I have an iPad app (IOS 4.3 & 5.0) that is creating a UIView with at least 2 subviews every time the user pages forward or backward. The 2 subviews are UIWebView objects.
On every relevant swipe the old UIWebViews are removed and deallocated properly (removeFromSuperview then stopLoading and set delegate to nil) and the parent UIView is deallocated. I can confirm that both parent UIView and the 2 child UIWebViews are absolutely gone.
I can see the retain count on the webviews as they are being removed and I am quite sure that they are indeed removed every time.
However my memory allocation continues to increase on every swipe by ~200-350k . No leaks in instrument but I can see the memory usage going up slowly.
On iPad 1 the app eventually (within 6-10 minutes of moderate usage) receives multiple memory warnings and is terminated. LowMemory crash log...etc is generated.
We use stringByEvaluatingJavaScriptFromString and loadRequest to populate the Webviews. Memory usage creeps up even when nothing is injected or loaded into the webViews.
Is anybody experiencing this kind of behavior with UIWebViews ? Has anyone dealt with this successfully ?
Thoughs, comments & answers would be greatly appreciated.
Try to use Instruments but with the "Allocations" template and watch the "# Living" columns for UIWebView. I usually use it like this:
Profile app with Allocations template
"Warn up" the app by going thru all tabs, scroll around etc.
Press "Mark Heap", this will create a "Baseline" heapshot
Do the thing you think causes objects to stay around
Press "Mark Heap" again, this will create a "Heapshop #" heapshot
Inspect the objects in "Heapshop #" which will show size and number of objects created and alive since the last heapshot.
Goto 4
Not aware of any leak, but if you are destroying and re-creating the same view hierarchy every swipe, why not store the views and re-use them instead?
Regardless of whether they leak, webviews are computationally expensive to set up, so re-using them and just reloading the content should give you a performance boost, and may solve your leak as well.
are you still seeing this issue? I traced something which looks like unbounded memory usage (not necessarily a leak, but I get low memory warnings and eventual process termination) by calling stringByEvaluatingJavaScriptFromString over and over:
[webview stringByEvaluatingJavaScriptFromString:#"something()"];
It seems like there is something odd there.