Re: [asm] Problem instrumenting EMMA-instrumented code

[Eugene Kuleshov <eu@xxxxxxxx>]

Luke Blanshard wrote:
> Actually I have a workaround.  I implemented visitLocalVariable to 
> check for the illegal state and simply discard the entry in that case.
 Which is the same as compiling class without the debug info.
> >  I would have to see the bytecode created by EMMA to suggest 
> > anything. The way you describe it sounds like a bug in EMMA. The 
> > javap output you showed doesn't give me a clue what happens with the 
> > LicalVariableTable attribute in those classes.
> Really?  Here was the original class's local variable table from javap:
> ...
> As I said, this is unusual but I believe technically legal.  The spec 
> says that the start and end locations must both be instructions.
>
> Is there more information about the local variable table available in 
> the class file?
 Since position in the bytecode is 0, ClassReader creates a new Label 
for that position, but then Label's offset is set to -1. Why that 
happens I can't explain without stepping trough reading of that class file.

 regards,
 Eugene