How does Eclipse/IDEA/etc. debugger obtain the information about local variable values and watch expression values?

Question

I always thought that varibales are mapped to stack locations once your Java source is compiled; additionally, they may include the info about the variable names and their scope in classfiles, but that's optional AFAIK.

The question is - how do my Eclipse/IDEA IDEs allow me to set a watch expression containing the local variable name? To me, it's hard to understand :)

Answer 1

All those IDE base their debug services on top of the JavaTM Platform Debugger Architecture (JPDA) .
More precisely, they are all using the Java Virtual Machine Tool Interface (JVM TI) , which replaced from JDK5 the experimental JVMPI (Java Virtual Machine Profiling Interface).

That JVMTI includes a Watched Field function and a SetFieldModificationWatch function .

But that doesn't include local variable : there is actually an RFE opened since quite some time now: bug 4228507 .
So in the case of local variable, chances are some kind of BCI is used ( Byte Code Instrumentation ) (like the TPTP Eclipse module does ).