Shigeru Chiba

I understand that calling super() twice (at two places) is not valid according to the specification. However, if ASM and BCEL can generate a valid stackmap for that code, Javassist...

Although my memory might be incorrect, I believe the JVM specification (not Java lang. specification!) restricts what a constructor can do with uninitialized `this` before calling super(). But, yes, the...

I never say what you're saying is shady! Your report is very valuable information that helps to improve Javassist's stackmap generator. If the specification were very clear, the implementation would...

Let me ask one question. Is the output of `javap` you pasted above: https://github.com/jboss-javassist/javassist/issues/328#issuecomment-655207465 the bytecode that ASM generated? Or the bytecode by Javassist? It looks like the code by...

Sure. Javassist automatically constructs a stackmap from the bare bytecode. Please look at `javassist.bytecode.stackmap`. It first constructs a control flow graph and computes a data flow on that graph so...

Very close! Although your suggestion is not an exact one, it helped me lots. Thanks.

I do not have any plan for release. Maybe releasing 3.28 in (late) August?

I cannot see any good algorithm to determine whether these arguments are valid or not. Can you?

Ah, I intended to say that I didn't know how to implement your idea. Since the source code of the method body is not available, making `insertBeforeAndAfter` work for arbitrary...

It might be good to add a bit more comment. Javassist modifies the bytecode without its source code as if it *virtually* allows the users to modify the source code....