bleg: debugging tools

[gusl]

I would keep 2 versions of the source: the current one, and a previous one (updated at certain "checkpoints", when I am confident it works correctly), and run them in parallel.

I want tools that would help answer the following questions:
* Compared to the checked version, at which point (i.e. line number) in the execution does a variable get a different value? At which point does the program branch differently?

Also:
* provide a way for managing prints in the two programs simultaneously: prints that appear in one should appear in the other.
* you should get the same "random" numbers in both runs. Is there some way to rig java.Math.random()?

Any such tools for Java?

Comments

[ikeepaleopard]

Look up regression testing. Java random numbers can be seeded if you use a new random number object instead of the java.Math one.

[bhudson.livejournal.com]

http://weblogs.mozillazine.org/roc/archives/2006/12/more_about_ambe.html

[jbouwens.livejournal.com]

Do you use version control? If so, use the system's tagging feature to create your "checkpoints" of verified code.

I'm not aware of any tools that will tell you specifically where in a test-run a different branch/value occurs. All regression test systems I know of allow you to write a test, and specify an expected result. Running the test then passes or fails, depending on whether the actual result matches the expected result.

[jbouwens.livejournal.com]

Actually, thinking a bit further, pinpointing where a different branch occurs is non-sensical in code that changed. And if the code did not change, then the only reason why it would branch differently is if the code's input has changed.

[darius.livejournal.com]

try http://en.wikipedia.org/wiki/Delta_Debugging