Android Question SubExists, CallSub and Obfuscation

[Cyan]

I encountered a difficult-to-find error when compiling my app to release (obfuscated) when checking if a class implements a particular method with SubExists.

It appears that whilst CallSub works fine in obfuscated mode (e.g. CallSub(myObject, "mysub")) as the string "mysub" is translated, it appears that SubExists(myObject, "mysub") does not.

Consider the snippet below, with myobject as a custom class that has one method, "getTitle"

Private Sub Class_Globals
End Sub

Public Sub Initialize
End Sub

Public Sub getTitle as String
    Return "Hello World"
End Sub

Dim myobject as MyClass
myobject.Initialize

If SubExists(myobject, "getTitle") Then
    Log(CallSub(myobject, "getTitle"))
End If

This will log nothing.

If however, I change this slightly to the below, then I obtain something in the log:

Dim myobject as MyClass
myobject.Initialize

Log(CallSub(myobject, "getTitle"))

My question is, is this expected behaviour for SubExists versus CallSub when obfuscating?

I've attached a sample project. Running it in "Release" mode works fine, however, running it with "Release (obfuscated)" does not.

 

[edgar_ortiz]

Did you check:

Code Obfuscation

This tutorial is relevant for B4A and B4J. During compilation B4A generates Java code which is then compiled with the Java compiler and converted to Dalvik (Android byte code format). There are tools that allow decompilation of Dalvik byte code into Java code. The purpose of obfuscation is to...

www.b4x.com

 

[Cyan]

Did you check:

Code Obfuscation

This tutorial is relevant for B4A and B4J. During compilation B4A generates Java code which is then compiled with the Java compiler and converted to Dalvik (Android byte code format). There are tools that allow decompilation of Dalvik byte code into Java code. The purpose of obfuscation is to...

www.b4x.com

Thank you, I did have a read about the CallSub section, however I am using a static string so according to that the sub should not be renamed. However, when 'wrapping' it with a SubExists, it looks like it's renamed. I guess this is just a compiler quirk and I'll either have to rename over 60 methods or wrap my calls in a Try instead.

 

[Sandman]

It appears that whilst CallSub works fine in obfuscated mode (e.g. CallSub(myObject, "mysub")) as the string "mysub" is translated, it appears that SubExists(myObject, "mysub") does not.

Assuming this is correctly described, I can't imagine it would be impossible to replace the "mysub" in the SubExist with the obfuscated name also? Perhaps you should post a Wish and point to this thread?

 

[Erel]

The SubExists shouldn't matter.

 

[Cyan]

The SubExists shouldn't matter.

Thank you for your time Erel - my apologies, I am not sure what you mean?

 

[Sandman]

I'm not sure about the meaning either.

I'm thinking that perhaps it's some kind of ancient wisdom, like "it doesn't matter how big the banana is, the peel is always bigger".

 

[agraham]

Erel's description in that 'Code obfuscation' link is not 100% complete in describing the implementation of CallSub in obfuscation. Obfuscation is in fact renaming myClass.getTitle because it lacks an underscore. For this reason SubExists doesn't see it. However the compiler does a hack on CallSub if the Sub name is a literal string to implement an extra method in myClass that does a check on the unobfuscated name and if it matches calls the actual obfuscated implementation so the only solution to get SubExist to work is to add an underscore to the Sub names.

 

[Sandman]

Yes, that sounds pretty much like I've understood it. My point was that perhaps the same hack could be made to SubExist?

 

[Cyan]

Erel's description in that 'Code obfuscation' link is not 100% complete in describing the implementation of CallSub in obfuscation. Obfuscation is in fact renaming myClass.getTitle because it lacks an underscore. For this reason SubExists doesn't see it. However the compiler does a hack on CallSub if the Sub name is a literal string to implement an extra method in myClass that does a check on the unobfuscated name and if it matches calls the actual obfuscated implementation so the only solution to get SubExist to work is to add an underscore to the Sub names.

Thank you!! This is the bit that I was missing, that makes it much clearer as to what is going on under the hood.

In this case, rather than renaming 60-odd methods, I will just wrap the single CallSub in a Try, as the methods return something.

 

[agraham]

perhaps the same hack could be made to SubExist?

Having looked at the code - probably.

 

[Erel]

Thank you, I did have a read about the CallSub section, however I am using a static string so according to that the sub should not be renamed. However, when 'wrapping' it with a SubExists, it looks like it's renamed. I guess this is just a compiler quirk and I'll either have to rename over 60 methods or wrap my calls in a Try instead.

I meant that the SubExists will never have any effect on the compiler. It isn't the reason that the sub name is being obfuscated. I can't say why it isn't obfuscated without a small project that demonstrates it.

 

[agraham]

I meant that the SubExists will never have any effect on the compiler. It isn't the reason that the sub name is being obfuscated. I can't say why it isn't obfuscated without a small project that demonstrates it.

Repectfully I think you missed the point of the question. There is no problem here except that SubExists does not work when the method name being tested is obfuscated because, unlike CallSub, there is no provision in SubExists for testing the literal string against the original unobfuscated name.

 

[Erel]

No. No. No.

When the compiler decides that a sub name shouldn't be obfuscated, it keeps the original name. It doesn't change the CallSub string. You can see it in the generated Java code.
This means that if CallSub works then SubExists should also work.

 

[agraham]

This means that if CallSub works then SubExists should also work.

I have looked at the generated code for the above project. It behaves as I have described. SubExists does not work when that project is obfuscated. Perhaps there is a compiler bug if you expect SubExists to work in this case.

 

[Erel]

Let me check it...

 

[Erel]

You are correct that this case is a bit more complicated. It fails because getTitle is treated as a property. I was surprised that CallSub worked and SubExists failed. This is related to an optimization (the callsub method in the class). Change it to GetTitle and it will work.

 

[agraham]

You are correct

Will this be 'fixed' in a later B4A version or will it remain a 'feature'?

 

[Erel]

It will be fixed at some point.