Over the last few days I’ve created a launcher with a GUI to make starting the thread monitor easier. The command line was getting a little bit long: Every time, you had to append the boot classpath of the program to be monitored. The launcher also allows you to save configurations. That’s a big plus.
I guess I should also generalize the FileInstrumenter and add a GUI. Right now it bugs me a little that I still need to change the paths depending on whether I’m in my office or at home.
Today I found out again why Corky is my advisor and I’m his student, and not the other way around (not that I ever forget that). We were talking about my problems with the stack frames, and he suggested I do my instrumentations in a way that matches handwritten Java code. That way, it should be easier to verify.
Why didn’t that occur to me?