4.6 iOS 64-bit beta

Thanks, we were on 5.0.1, but 5.0.1p4 does seem to fix it.

 

@JoshPeterson

Hi Josh,

Did anyone look at issue 689976? We're still getting the deadlocks with 4.6.5 and the issue has been set to Closed with no comments.

Thanks!

 

Thanks for your and the Unity team's hard work. I'm ok with these warnings as long as I know they're benign. I'm sure that soon everything will be back to normal. Thanks for putting my mind at ease about these warnings.

 

@JoshPeterson

Just tried it with 4.6.5p1 and it locked up on my first 5 attempts before eventually settling in at an ~60% repro rate. Running the Release scheme in XCode seems to increase the likelihood of it occurring. I'll also email the bug report with this info.

Thanks for the continued efforts!

 

@JoshPeterson

Bug 689976. Thanks Josh!

 

Calling function via a delegate to an interface broken again. Last working version is 4.6.4p1.
This bug catch in real game, but I can't reproduce them in small test project.

EDIT: Yeah! I reproduce this! Simple wrap test in template class!

Code (CSharp):

1. using UnityEngine;
2.  
3. interface ITest1 {
4.     void Test1();
5. }
6.  
7. class TestClass<T> where T : ITest1 {
8.     public string DoTest(T t) {
9.         System.Action a = t.Test1;
10.         return a.Method.Name;
11.     }
12. }
13.  
14. class Main : MonoBehaviour, ITest1 {
15.     public void Test1() {}
16.  
17.     void Start() {
18.         var c = new TestClass<Main>();
19.         Debug.Log(c.DoTest(this));
20.     }
21. }
22.  

 

Hello, can you help me? im trying to build my game for IOS64 and build operation works fine. But when i run my game on device game crashed. Mono version works fine.
Log:

FormatException: Input string was not in the correct format

at System.Type.GetTypeFromHandle (RuntimeTypeHandle handle) [0x00000] in <filename unknown>:0

at System.Int32.TryParse (System.String s, Int32& result) [0x00000] in <filename unknown>:0

at GlobalParser.ParseSingleValue (System.String source) [0x00000] in <filename unknown>:0

at GlobalParser.ParseCondition (System.String source) [0x00000] in <filename unknown>:0

at QuestsDataBase.Parse (System.String input) [0x00000] in <filename unknown>:0

at QuestsDataBase.Initialize (Int32 campaignIndex) [0x00000] in <filename unknown>:0

at DataBase..ctor (Int32 campaignIndex) [0x00000] in <filename unknown>:0

at GameCore.Create (Int32 campaignIndex) [0x00000] in <filename unknown>:0

at MenuLoader+<NewGame>c__Iterator83.MoveNext () [0x00000] in <filename unknown>:0

at System.Array+InternalEnumerator`1[UnityEngine.Color].get_Current () [0x00000] in <filename unknown>:0


(Filename: currently not available on il2cpp Line: 4294967295)

 

Sure I can. But since it occurs only at runtime, I'm afraid to miss something.

 

With regards to size again. I noticed that the aot compiler generated method and type names in the .s files as m1... or t1.. and conversely in the il2cpp it generated a really long method name.

A simple analysis of the resulting il2cpp binary file shows that the link segment and the text segment is almost the same size.. meaning, symbol names are just as big as my code size. Now that means that we also have many functions that do very little so much so our method names are bigger than the actual generated code..

Was wondering if there was a simple switch that can be turned on so that the il2cpp generated method names and types could be smaller or shortened..

 

Apologies.. I just saw that there is a command line switch
–output-format=Compact
Where/how can I switch it on to test?

 

Latest 4.6.5p1 il2cpp compilation is crashing with too many heap sections for our game. 4.6.4p4 compiles fine.

 

I'm not using development build, it's definitely not there 4.6.4p4
4.6.5p1 is crashing so I can't really tell unfortunately

 

I got a crash caused by dynamic batch in Unity4.6.4p4. (GfxDevice :: DynamicBatchMesh crash)
Now as I know, this crash only occurred on iPhone5S with iOS 7.0.1, the crashed ipa was built by IL2CPP, with dynamic batch turned on ( default setting ).

Mono(2.x) build is fine.
IL2CPP build but turned off dynamic batch in build-setting-panel is fine.

I do believe this is a IL2CPP bug in Unity4.6.4p4, could you got anything useful from the log?
Please check the attach. Thanks.
@JoshPeterson

 

4.6.5p1 crashes when building under windows. No problems on OSX.

 

Mantas. No, I was inspecting the .app that was produced from a released build. And yes, I just only figured out that symbols weren't stripped at all, during a normal release build with all optimizations switch on. I've managed to manually strip the symbols and it's much much better. I guess having -output-format=Compact doesn't have much impact on the code size with all symbols stripped out then.

What about –enable-generic-sharing? I checked using process explorer and it doesn't seem like it was on the command line for il2cpp.exe. Is that automatically turned on? If not, can I switch it on? This was in reference to Josh's blog on il2cpp internals.

 

We're using 4.6.5f1 for ARM64 IL2CPP building. We use System.Net.Sockets for non-blocking TCP connecting. Things work well in mono backend. But with IL2CPP backend, when socket.connect() called, an socket exception with native code 10036(InProgress) raised. We're sure this is the only socket we created.

Code (CSharp):

1.     private bool connect()
2.     {
3.         try
4.         {
5.             socket = new Socket(
6.                 AddressFamily.InterNetwork,
7.                 SocketType.Stream,
8.                 ProtocolType.Tcp);
9.            
10.             socket.NoDelay      = true;
11.             socket.Blocking     = false;
12.            
13.             socket.SendTimeout      = -1;
14.             socket.ReceiveTimeout   = -1;
15.            
16.             socket.LingerState      = new LingerOption(true, 0);
17.            
18.             socket.Connect(ipEndPoint);
19.         }
20.         catch (Exception e)
21.         {
22.             SocketException socketException = e as SocketException;
23.             if (null == socketException || SocketError.WouldBlock != (SocketError)socketException.ErrorCode)
24.             {
25.                 Debug.LogError(string.Format("BaseSocketTask Connect Exception: {0}", e.Message.ToString()));
26.                
27.                 return resultCallback(SL_LOCAL_NETWORK_CONNECT_FAIL);
28.             }
29.         }
30.        
31.         statusStartTick = Time.realtimeSinceStartup;
32.        
33.         return true;
34.     }

When socket.Blocking set to true, the connection is succeed and no exception raised. Non-Blocking is key to our game. Any advice?

 

Submitted a crashing bug with repro project http://fogbugz.unity3d.com/default.asp?697063_hrfvt3jr6qujsi3m

 

We see that when building il2cpp universal the Build Active Architecture Only settings is set to Yes for Debug scheme. Is this correct? As I understand, for Debug builds executable will only contain arch of connected device. Also, if I understand, memory consumption of debug build wil be different (smaller) than on release build. We easily can set this setting manually or in postprocess script to No. But maybe its correct to set this setting on Unity level when building il2cpp Universal?
Sorry for my English

 

Any progress on version 4.6.5p2?

 

Hi Josh. We also have a problem with the deadlocks. It also happens in 4.6.5p1.
Here - a screenshot in xCode maybe it helps.

 

Any update on the iOS il2cpp binary size problem? Especially with Universal architecture. It's not unreasonable that universal architecture takes twice the size, it's just very hard (practically impossible) to stay below apple's 100mb limit.
How much of our code goes into that one main binary and can we do anything to move it somewhere else?

It was announced at 4.6.4p1 that you did improvements to the size, should we expect any additional reductions in the near future? With 4.6.5p1 we're still quite far above the limit. Moving the code to a dll or something would probably solve our immediate problem as that would allow it to be properly compressed.

 

Ok, thank you for the update.

 

Thanks, I had missed that thread, but while it was very illuminating it doesn't solve our problems. Glad to hear build size is still highest priority.

 

Has anyone else found that having a script with OnCollisionEnter() without the Collision parameter crashes the app when the collision occurs with IL2CPP? I've submitted a bug report, number 697738.

 

What version of Unity is that with? I seem to remember some fix about crashes when event parameters didn't match on some patch release.

 

5.0.2p1. I haven't tested with any other version.

 

@JoshPeterson 4.6.5p2 solved the bug we had, but I sent another different repro project because there is a compilation error which seems to be related with generics. It used to compile in 4.6.5p1.

 

Yay, some good fixes/improvements in there.

Only you changed the stripping to be even more aggressive, my link.xml file is growing.

Let's see it was a static readonly dictionary with a constructor initialiser calling a generic static method with arguments generated by an anonymous method. Good luck figuring that one out. (NodaTime library)

Edit: Oups forgot to mention that the anonymous method returned an anonymous struct thingy. Want me to send you the entire static constructor code block, I'm sure that's enough to get the picture.

 

Hi there, I'm using 4.6.5p1 and I upgraded an old project to use IL2CPP. Compilation works fine but I noticed that on old devices (iPad 3, iPad Mini 2nd generation) the app now takes 20-30 seconds to launch (was taking 15s before). The good news is that on iPhone 6 Plus the app launches in only 4 seconds (THAT'S FAST!). Is there a way to optimize old devices launch times or is it due to the IL2CPP compilation and can't be improved? Thanks for letting me know.

 

If I change the Run scheme to release the app fails to launch for some reason. I usually never saw big changes in launch times between apps deployed on device through XCode and those deployed through HockeyApp of Testflight. Should I still file a bug report? As I mentioned on new device the app launches super quick. Thanks for letting me know.

 

The compile error bug report I encountered can be found here http://fogbugz.unity3d.com/default.asp?697757_rb4pcgk9h98n8t3c