I'm adding several UIView
objects (e.g. 5) to the screen, one inside the other. This, for example, view5.superview = view4
, view4.superview = view3
, view3.superview=view2
, view2.superview = view1
. For all these UIView
I set uitapgesturerecognizer; for view1-4 I just do NSLog(@"tap %@", self) in callback, while for view5 tap I set the following: delete view4 from the hierarchy, then put the same object view4' at the same place of the hierarchy. This object also contains view5' for which UITapGestureRecognizer
is set (practically, I replace one part of markup with similar one).
Then I start clicking on view5. Some time view5 keeps catching its tap and everything's OK, but random number of taps later (every time this number is different) one of the view1-4 objects starts catching this tap, though we're still clicking the view5. The whole problem has a random character - sometimes it occurs at the 10th launch, sometimes at the second. Sometimes wrong objects start catching taps at the first tap. Also I never know what object will catch a tap, when everything goes wrong. The frame for view(n+1) was set, e.g., as a half of the frame view(n), while the frame for view1 - e.g. (0,0 320, 460).
All operations with ui objects described above are conducted in the main thread, and everything I've told about perfectly worked on iOS 4.3 - 6.1 with much more complex examples. But the iOS7 makes out of it some kind of a random hell.
Update: I've created a sample project, to simplify the debug process. No add/remove subview operations on tap. Only 4 views on screen, on tap the app logs what view was tapped. So, you need to tap on smallest view (4). If you see "tap 4 tap 4 tap 4…" in the log - this is the case when everything works fine, stop and run again, stop and run again, stop and run again, etc. And at some runs (maybe after 10+ successful runs) you won't see "tap 4" on the first line, you will see "tap 1" or "tap 2" or "tap 3", and it will continue so - these are the bad cases.
Sample project can be downloaded from here: http://tech.octopod.com/test/BuggySample.zip (just 33 Kb in archive).
Update 2
We've posted a bug to Apple, I'll post here when we will get some feedback. However, any good workaround would be much appreciated!
Update 3
Solution, provided by Yuvrajsinh is really working on the sample project. Unfortunately, it still does not help to solve the problem that occurred in the main project where it initially appeared. The main reason for now is that if any view without self gesture is laying upon the clickable content, random view element under it starts catching the interaction (instead of the top one with interaction gesture set. Do you have any ideas how it can be solved? The updated sample can be downloaded from here: http://tech.octopod.com/test/BuggySample2.zip
Because the problem is only occurring in iOS 7, you can use one of the new delegate methods to resolve the issue:
– gestureRecognizer:shouldRequireFailureOfGestureRecognizer:
– gestureRecognizer:shouldBeRequiredToFailByGestureRecognizer:
I resolved it by implementing gestureRecognizer:shouldBeRequiredToFailByGestureRecognizer
and "crawling" up the gesture's view's superview so I could return "YES" if I find the superview's gesture is equal to the one provided. I detail my full resolution here: https://stackoverflow.com/a/19659848/1147934.
Explanation
The problem with gesture recognizers in iOS 7 is that a superview's gesture is receiving its touches before one of its subview gestures receives its touches. This causes the superview gesture to recognize which then cancels out the sub view's recognizer... this is (incorrect?) and multiple bugs have been filed with Apple. It's been pointed out that Apple doesn't guarantee the order in which gestures receive touches. I think a lot of "us" have been relying on an implementation detail that changed in iOS 7. This is why we use the new delegate methods, which seem designed to help us address this problem.
Note: I did extensive testing by using my own sublcassed recognizers, logging all touches and discovered that the reason recognizers fail is because superview gestures were receiving touches before a subview's gesture was in about ~5% of the cases. Every time this happened, failure occurred. This does happen more often if you have "deep" hierarchies with lots of gestures.
The new delegate methods can be confusing, so you need to read them carefully.
I'm using the method (I've renamed the arguments to make them easier to understand)
– gestureRecognizer:(UIGestureRecognizer *)thisRecognizer shouldBeRequiredToFailByGestureRecognizer:(UIGestureRecognizer *) otherRecognizer
.
If you return "YES", then the gesture recognizer provided, otherRecognizer
, will require thisRecognizer
to fail before it can be recognized. This is why, in my answer, I crawl up the superview hierarchy to check if it contains a superview that has the otherRecognizer
. If it does, I want otherRecognizer
to require thisRecognizer
to fail because thisRecognizer
is in a subview and should fail before it's superview's gesture is recognized. This will make sure that subview gestures are recognized before their superview's gestures are. Make sense?
Alternative
I could go about it the other way around and use:
– gestureRecognizer:(UIGestureRecognizer *)thisRecognizer shouldRequireFailureOfGestureRecognizer:(UIGestureRecognizer *)otherRecognizer
Now I would need to crawl through my entire subview hierarchy, check if otherRecognizer
is in it and return YES
if it is. I don't use this method because crawling the entire subview hierarchy is much more difficult and expensive to do than to check a superview hierarchy. Crawling a subview hierarchy would have to be a recursive function, while I can use a simple while
loop to check a superview's hierarchy. So I recommend the first approach I outline.
Warning!
Be careful about using gestureRecognizer:shouldReceiveTouch:
. The issue is a problem of which gesture receives touches first (canceling the other gesture out)... it's a problem of conflict resolution. If you implement gestureRecognizer:shouldReceiveTouch:
, you risk rejecting a superview's gesture if the subview gesture fails because you have to guess when a subview gesture might be recognized. A subview gesture may legitimately fail for reasons other than the touches are out of bounds, so you would have to know implementation details in order to guess correctly. You want the superview gesture to be recognized when the subview gesture fails but you don't really have anyway to know for certain if it will fail before it actually fails. If a subview gesture fails, normally you want the superview gesture to then recognize. This is the normal responder chain (subview superview) and if you mess with that you could end up with unexpected behavior.