Each method has its strengths. The warning handler takes less work to do. But will break if anyone else decides to create a warning handler. And it doesn't help if you trying to trace what happened through 3 eval statements in a long stack backtrace. (Most of my errors already use Carp::confess, so petdance's proposal doesn't add anything that I already usually have.)
The alternative that I suggested takes more work, but adds context to the error message. Furthermore it isn't any more work than you have to do already just to undo the fact that eval traps errors.
I find my alternative particularly useful when the eval constructed a function that is being called dynamically later. (Like the example that I pointed at in my code.) In that case a stack backtrace only tells you that method X was called somewhere, and that method X was generated in some eval, in some class. Quick, which class was it produced in, and find me the code that produced it! | [reply] |