4.6 iOS 64-bit beta

Hi Josh, has there been any progress on this issue with generated xcode project from last week? Bug # # 688073 On a positive note, our other app that we have in development on 4.6.x was able to build to il2cpp without issue and is running well, just need to get this one rolling.

Thanks,
RJ

 

We are using Unity 5.0.1.
When we compile for il2cpp we get "semantic issue" errors. This semantic issues come from the photon plugin, but it didn't happen before. Is there something we need to do so there are no semantic issues when unity makes the build?

 

Thank you for such a quick response @JoshPeterson. I tried the latest patch and the semantic issue problem is gone. Seems like it was only ocurring in 5.0.1. We'll run more tests to triple check.

 

@JoshPeterson

The bug report just finished uploading. It's Case 689976, let me know if you have any issues reproducing it. Thanks!

 

@JoshPeterson

Last month, we reported a bug (case 682047), and also we received a email from your QA Team.
But till now, we dont know any progress of this bug,Would you please give me any way to follow this bug? Or Is there any update about this bug.

ps:
this bug is about a sqlite problem in the IL2CPP under unity4.6.x. we report this bug use 4.6.3f1 ,and we tested under 4.6.4f1,the bug still not fixed!

 

@JoshPeterson
Good morning ,

right now i am using Unity 4.6.3.p3

Unity 4.6.4 may fix this problem --> iOS/IL2CPP: Corrected the C++ compiler error in generated code: "error: cast from pointer to smaller type 'il2cpp_array_size_t' (aka 'int') loses information" .

i think my problem is solved by unity 4.6.4 already.
thnakx a lot Mr. Peterson..u helped me very much.. if my problem comes again i will disturb u again.

 

There have been major changes and fixes in IL2CPP in the 4.6.4 patch releases, I would try with the latest patch release, just in case your bug is already fixed.

 

@JoshPeterson
After updating to version 5.0.1.p1, I got a lot of errors, such as:
Classes/Native/Il2CppMethodTable.cpp:7:17: Conflicting types for 'List_1__ctor_m10644_gshared'
Classes/Native/Il2CppMethodTable.cpp:13:17: Functions that differ only in their return type cannot be overloaded

I tried to compile empty clear project, but got the same errors.

Folder Classes/Native in version 5.0.1.p1 takes ~ 60 mb, but in version 5.0.1.f1 ~ 120 mb, apparently it affects the result.

Now every new update it as an entirely new system, do not know what new bugs appear.

 

Can we expect a 4.6 patch release today?

 

We are using Crashlytics for crash reporting. In the past we've gotten a nice stacktrace about crashes. When I updated to Unity 4.6.4 and enabled IL2CPP and the Universal 64-bit mode, the stack traces are as follows:

Code (csharp):

1. 0 IL2CPP  il2cpp::vm::Runtime::GetVirtualInvokeData(MethodInfo*, void*)
2. 1 IL2CPP GeneratedVirtualInvokers.h line 8 ActionExtension_SafeInvoke_m8383(Object_t*, Action_t120*, MethodInfo*)
3. 2 IL2CPP  GeneratedVirtualInvokers.h line 8 ActionExtension_SafeInvoke_m8383(Object_t*, Action_t120*, MethodInfo*)
4. 3 IL2CPP  GeneratedVirtualInvokers.h line 9 EventDelegate_Execute_m3911(EventDelegate_t849*, MethodInfo*)
5. 4 IL2CPP  GeneratedVirtualInvokers.h line 19 EventDelegate_Execute_m3914(Object_t*, List_1_t765*, MethodInfo*)
6.  

@JoshPeterson Any ideas what we are missing? How can we get the proper class and method names back to the crash reports?

Edit: Added code formatting.

 

It's out!

 

Hello,

I tried upgrading our tool stack from

Unity 4.6.3f1 / Xcode 6.0.1
Unity 4.6.4f1 / XCode 6.2

(IL2CPP)

And I am seeing this crash. I ALSO tried 4.6.4p2.

Known issue ? If not I'll report it.

I'll now try p3 and different combinations of Unity3d / Xcode.

 

@JoshPeterson I'm expecting a stack trace that contains the name of the class and the method where the error occurred. However, since I posted the message I've started to suspect that this could be an IL2CPP bug instead?

Here's how another crash looks like. In this case, I'm triggering the crash myself from a debug panel.

Code (csharp):

1. 0  IL2CPP                         0x0000000101840c70 -[Crashlytics crash] + 12
2. 1  IL2CPP                         0x0000000101840c70 -[Crashlytics crash] + 12
3. 2  IL2CPP                         0x00000001002860d4 DebugManager_OnGUI_m5431(DebugManager_t1102*, MethodInfo*) (Bulk_Assembly-CSharp_6.cpp:12244)
4. 3  IL2CPP                         0x0000000100c02584 RuntimeInvoker_Void_t22(MethodInfo*, void*, void**) (GeneratedInvokers.cpp:12)

The key here is that I can see "DebugManager_OnGUI" which tells me where the problem occurred!

Just look at the stack trace I pasted in my previous comment. There is absolutely no hint as to what is wrong - none of the lines contain any references to our classes or methods. How am I supposed to fix such a crash?

 

Actually, it seems that the "ActionExtension_SafeInvoke" and "EventDelegate_Execute" are in fact real methods in our (3rd party) code. Apologies. They sounded so generic that I thought they were related to IL2CPP.

However, the original issue remains:
1) I don't see this crash in the editor (as a warning nor error)
2) This crash does not occur when compiling with Mono

How do I go about fixing this crash?

Also the stack trace still looks very weird to me. EventDelegate.Execute() does exist but the code doesn't know anything about ActionExtension.SafeInvoke(). How can it be calling it? Are the IL2CPP stack traces possibly missing calls to anonymous methods? Are the method contents possible inlined somehow? How can I know what is the actual class and method where this error occurs?

 

After downloading and installing 4.6.4p3 OSX and Unity say the version number is 4.6.4p2. I also tried uninstalling Unity beforehand. Anyone else experiencing this?

 

@JoshPeterson

The original stripping level was micro mscorlib.

• Disabling stripping on IL2CPP fixes it
• Switching to Mono backend with the original stripping level fixes it
• Reverting to 4.6.3f1 with IL2CPP and original stripping level fixes it

Note: the issue is on a scene that uses the camera. There have been Camera related changes in 4.6.4p1 (and maybe in 4.6.4 as the p1 was supposedly a fix for a regression).

Update:

• 4.6.4p3 also affected
• sending issue right now

 

Is there something akin to the il2cpp_extra_types.txt file for methods? I'm currently getting the following error in my IL2CPP project:
System.NotSupportedException: Failed to construct generic method 'System.Reflection.MonoProperty::GetterAdapterFrame' with generic arguments [Mono.Security.Protocol.Tls.HttpsClientStream, System.Boolean] at runtime.

Previously, I had:
System.NotSupportedException: Failed to construct generic type 'System.Reflection.MonoProperty+Getter`2' with generic arguments [Mono.Security.Protocol.Tls.HttpsClientStream, System.Boolean] at runtime.

I found a post that pointed me to the il2cpp_extra_types.txt file to force IL2CPP to force compile the Getter class. After I did that, that error disappeared and I got the other one. Is there something I can do to force compile that generic method? Thanks

 

@jerome.lacoste

This sounds very much like another manifestation of 684129.

The workaround of disabling stripping doesn't have a huge impact on size, as IL2CPP always strips anyway. Alternately, you can find out what class it is trying to awaken and add initialisation for it in UnityClassRegistration.cpp.

 

Any new updates on getting GetManifestResourceInfoInternal implemented?

 

In 4.6.4p2 and p3, I have some problems with Dictionaries. I have something like:
private Dictionary<int, Touch> touchPoints = new Dictionary<int, Touch>();
voidUpdate () {
if (touchPoints.ContainsKey(...)
and get a bad access exception in:
il2cpp::icalls::mscorlib::Mono::Runtime::mono_runtime_install_handlers:
The same method uses another dictionary from int to bool which works just fine.

Any thoughts? Any place where I can report this?

 

@JoshPeterson
Could you please check this log? it's a crash of dynamic batch.
I just get this log from QA, not located where it happen yet.
Do you know this issue?
(Unity 4.6.4p2, IL2CPP)

 

@JoshPeterson

m very disappointed 4.6.4 resolve the problem of il2cpp_codegen_marshal_allocate_array .
but my problem is still there.

I add a log on player position .. Xcode prints my player position is a NaN number on device but the position is correct on unity editor.. m very confused how the things are going on..i have no idea.

 

Hi Josh,

Just an update, p3. Still doesn't resolve my issue 689022. This is just a confirmation. That is all. Thanks.

 

@JoshPeterson

aftercare research i found that 4.6.4 resolve the errors of il2cpp_codegen_marshal_allocate_array.

but there are still 370+ warning related to value conversion issue.
I think the IL2CPP built lib (Bulk ICSharpCode.sharpZipLib_0.cpp or other) does not full fill the requirement of app store . According to App store 64 bit requirement

int *pointer = some value;

int *next_pointer = (int *)((int) pointer + 1); // Incorrect

int *next_ pointer = pointer + 1; // Correct

or many more other rules.
plz visit the page
http://www.chupamobile.com/blog/2015/01/19/convert-app-64-bit-requirement/

In BulkIcSharpCode.sharpZipLib_0

return (((int32_t)(*(uint32_t*)(uint32_t*)SZArrayLdElema((((Crc32_t943_StaticFields*)InitializedTypeInfo(&Crc32_t943_il2cpp_TypeInfo)->static_fields)->___CrcTable_1), L_0))^(int32_t)((int32_t)((uint32_t)___oldCrc>>8))));

i think which is not fulfill the requirement of App store 64 bit requirements. so, may be this cause the warnings.

hereby i attach a snapshot of all the classes which have warnings.

 

i'm new here, and i just find many problems when building the standard example of unity 5.0.1 to xcode project. build for the simulator and got many warnings,just see nothing at the ipad simulator.does anybody know how to put the example run on ios.

 

@KEVINWEN
i think 5.0.1 does not work on simulators (not sure about it) try it on device

 

First post of this thread:

 

Im seeing the same issue here too. Im still running into the same issues as p2 so i cant confirm if its actually just a install label bug.