Class 30: Proving Programs Correct

Held Tuesday, March 14, 2000

Overview

Today, we consider one way that we might make programs more reliable: by proving them correct.

Question 30 for today's class: What do we mean when we say a program is "correct"?

Question 31 for Wednesday's class: How do we balance the need for formal specifications required by program proof techniques and the complexity of modern graphical applications?

Notes

• On Friday, we'll work on BillMan. You may want to start thinking about what features you'd like to see in BillMan.

Contents

• When is a program correct?
• Proving Programs Correct
• Applicability of Program Verification

Summary

• What does it mean that a program is correct?
• How do we prove that a program is correct?
  • Example (from the book): GCD
• Can we show that a nontrivial program is correct?

When is a program correct?

• We begin with the question I asked you for today: What does it mean when we say that a program is correct?
• There are many different perspectives on this issue. We'll look at some.

Proving Programs Correct

• To many mathematically-oriented folks, a program is correct when we can prove that it does the correct thing
  • Most typically, that the output is correct for every reasonable input
• I'm assuming that most of you did not successfuly follow the proof in the textbook, so we'll try it.
• We may also do a more formal proof for binary search.

Applicability of Program Verification

• Can we apply this technique to all programs? Why or why not?