Option to not inline inner classes?

Hi, I’m the dexpatcher user that started the discussion in https://github.com/DexPatcher/dexpatcher-tool/issues/33

Thanks so much for the patch made in response to that!

I’m testing out a build from master with that, hoping to compare the difference with the new checks vs --usesignatures however I’m running into some decompilation failures with other parts of the app.

The new issue seems to be caused by an inner class (obfuscated to the name a06$a) being inlined. One of this inner class’ functions then tries to create an instance of itself and cfr responds with ig62 = new /* invalid duplicate definition of identical inner class */;

I more often work with jadx with the arg --no-inline-anonymous to keep inline classes in their separate named form. The decompilation from that looks a little like:

public final class a06 extends td {
    public static final class a extends sf6 implements ig6<il6, xe6<? super cd6>, Object> {
        public final xe6<cd6> create(Object obj, xe6<?> xe6) {
            a aVar = new a(this.this$0, this.$block, xe6);
        }
    }
}

However cfr is inlining the inner class a into the outer class function that uses it.

public final class a06 extends td {
    public final rm6 a(hg6<? super xe6<? super List<DebugFirmwareData>>, ? extends Object> hg62) {
        return gk6.b((il6)jl6.a((af6)zl6.a()), null, null, (ig6)new ig6<il6, xe6<? super cd6>, Object>((a06)this, hg62, null){

            public final xe6<cd6> create(Object object, xe6<?> ig62) {
                ig62 = new /* invalid duplicate definition of identical inner class */;
            }
        }
    }
}

Is there an arg for CFR to not inline inner classes? I’ve tried a lot from --help but haven’t found anything that works.

[andrewleech]

Hi, no pressure on backlog I know exactly what it’s like!

I seem to have crafted a test case that represents it, I think the key features are the inner class with a function that needs to create another instance of the same inner class. jadx:

package com.test;

public final class outer {

    public static final class inner {
        public final Object create(Object obj) {
            return new inner(this.this$0);
        }
    }

    public final Object func() {
        return new inner(this);
    }
}

cfr:

/*
 * Decompiled with CFR 0.150-SNAPSHOT.
 */
package com.test;

public final class outer {
    public final Object func() {
        return new Object(this){

            public final Object create(Object object) {
                return new /* invalid duplicate definition of identical inner class */;
            }
        };
    }
}

jar attached 😃 testcase-inner-recursive.zip

[andrewleech]

Very interesting, I wonder how the original code was generated this way! Thanks for the added options though, that sounds perfect. I’ll give them a try next time I rework my project!