1 / 51

Disciplined Software Engineering Lecture #13

This lecture covers trace tables, symbolic execution, case checking, and program proofs for design verification.

garydsmith
Télécharger la présentation

Disciplined Software Engineering Lecture #13

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Disciplined Software Engineering Lecture #13 • Software Engineering Institute • Carnegie Mellon University • Pittsburgh, PA 15213 • Sponsored by the U.S. Department of Defense

  2. Design Verification, Part 2 • This is the second lecture on design verification. • It covers • trace tables • symbolic execution • case checking • mathematical program verification • program proofs

  3. Trace Tables • Trace tables are similar to execution tables, but more general. • Trace tables examine general program behavior rather than verifying individual cases. • Trace tables use • symbolic execution • case checking

  4. Symbolic Execution - 1 • In symbolic execution • the program is represented symbolically • program behavior is examined analytically • Symbolic execution can produce quite general verifications. • Symbolic execution is not always practical.

  5. Symbolic Execution - 2 • In symbolic execution, the approach is to • assign algebraic symbols to the program variables • restate the program as one or more equations in these symbols • analyze the behavior of these equations

  6. Symbolic Execution - 3 • Some questions to ask are • does the program converge on a result? • how does the program behave for both normal and abnormal input values? • does the program always produce the desired results?

  7. Symbolic Execution Example • begin • V1 = V2 • V2 = V4 • V3 = V1 • V4 = V3 • end • Given the following simple program segment • What would be the result for various values of V1, V2, V3, and V4?

  8. The trace table for this example would look like this: • Thus the result, for inputs A, B, C, and D is B, D, B, and B. V1 # Instruction V3 V4 V2 A C D B Initial values 1 V1 = V2 B 2 V2 = V4 D 3 B V3 = V1 4 B V4 = V3 Final values B B D B

  9. Symbolic Execution Exercise • Determine the values for which the following program will terminate. • Calculate(x: real) • if x>0 • x=sqrt(x)-1 • else • x=-x • if x<>0 • Calculate(x) • end.

  10. Exercise Result - 1 • The variables can be represented by x1 and x2. • The equation for the result of each cycle is • if x1>0: x2 = sqrt(x1) -1 • if x1 <= 0: x2 = -x1

  11. Exercise Result - 2 • For x1 = 0, the program terminates on the first cycle. • For x1 = 1, the program terminates on the first cycle. • For x1 = -1, the program terminates on the second cycle.

  12. Exercise Result - 3 • Working backwards • for x2 to be 1, x1 would have to be 4 • for x2 to be 4, x1 would have to be 25 • for x2 to be 25, x1 would have to be 676 • for x2 to be 676, x1 would have to be 458329 • and so forth • Thus, the program terminates for input values of + or - 1, 4, 25, 676, 458329, etc.

  13. Exercise Result - 4 • For other large positive or negative values of x1, the value of x2 will converge to a non-zero number between -1 and +1. • What happens for non-zero values between -1 and +1?

  14. Exercise Result - 5 • Pick x1=a, a very small number • x2=sqrt(a)-1, a negative number • next, x2 is made positive • in 2 more cycles, x2 is approximately sqrt(a)/2, or slightly larger than a • For small x1, the value of x2 thus gradually grows larger than x1. • The value of x2 will thus not reach 0 but will converge on some number less than 1.

  15. Exercise Result - 6 • When the value converges, x1 = x2 = x and the result would be given by the equation • sqrt(x) - 1 = -x • the solution to this equation is • x = [3 - sqrt(5)]/2 = 0.381966 • Thus, the program will end up cycling between the values of + and - 0.381966.

  16. Symbolic Execution Advantages • Symbolic proofs can be general. • Symbolic proofs typically involve less work than other methods. • Symbolic proofs are less error prone than executing test cases.

  17. Symbolic Execution Disadvantages • Symbolic execution is hard to use except for algorithmic and substitution problems. • Symbolic execution proofs are manual and thus error prone. • Symbolic execution is difficult to use with complex logic.

  18. Case Checking - 1 • Case checking involves the following steps • 1. Examine the program and determine the • various categories of program behavior. • 2. For these categories, identify the cases to be • checked. • 3. Check the cases to ensure that they cover all • possible situations.

  19. Case Checking - 2 • 4. Using trace tables, either alone or with • symbolic execution or proof by induction, • examine program behavior for each case. • 5. Once every case has been verified, program • behavior has been verified.

  20. Trace Table Example - 1 • The ClearSpaces program requirements • clears the leading and trailing spaces from a character string • leaves the embedded spaces in place • produces a state value 0 for an empty string • produces a state value 1 for a 1 space string • produces a state value 2 for a 2 or more space string • produces a state value 3 for a string containing non-space characters.

  21. ClearSpaces(char Input[ ], int State) {Length = strlen(Input); if (Length > 0 ) while(State!=3 && Length!=0) if (Input[Length-1] = ‘ ‘) {Length = Length - 1; if (State < 2) State=State+1; } else State = 3; Trace Table Example - 2 • The portion of ClearSpaces for removing trailing spaces is

  22. Trace Table Example - 3 • To define the cases for trailing spaces, establish the following conventions • 1. A message string is a string starting and • ending with a non-space character. • 2. An input string will consist of m message • characters and t trailing spaces • 3. m is either >0 or 0 • 4. t is either >0 or 0

  23. Trace Table Example - 4 • There are thus a total of 4 cases • 1. 0-0: m = 0, t = 0 • 2. 0-t: m = 0, t > 0 • 3. m-0: m > 0, t = 0 • 4. m-t: m > 0, t > 0

  24. Trace Table Example - 5 • The strategy is to verify program behavior with each case, using whatever method is most appropriate. • 1. Check the 0-0 case by itself, using an • execution table. • 2. Check the other cases, using proof by • induction. • 3. Do each proof with a trace table, using • the execution table method described in • lecture 12.

  25. Verifying Program Correctness • These methods are based on treating programs as mathematical theorems. • They are particularly useful during design and design review. • Because program verification often involves sophisticated reasoning, it is important to check your work carefully.

  26. Correctness Verification • The correctness verification approach is to • identify all the program cases. • use trace tables to evaluate all the conditions the program must satisfy for every case • use proof by induction where appropriate • The program is verified if all the conditions are satisfied for every case. • While this verification approach works with most design constructs, only loops are described in this lecture.

  27. ForLoop Verification - 1 • The ForLoop is assumed to have the following form • ForLoop • for n = First to Last • begin • nPart • end

  28. ForLoop Verification - 2 • The for-loop conditions require that the answer to the following question be true • Does ForLoop = FirstPart+SecondPart+...+LastPart

  29. ForLoop Example - 1 • Verify that the following program produces the factorial of n. • long factorial(int n) • { • x=1; • for (i=1; i <=n, i++) • x = x*I; • return x; • }

  30. ForLoop Example - 2 • First, identify the cases. • The cases are for n = 1 and for n > 1. • For n = 1 • ForLoop = 1! = 1 • FirstPart+SecondPart+...=FirstPart = 1*1=1 • The case for n = 1 is thus valid.

  31. ForLoop Example - 3 • For n > 1, say 3 • ForLoop = 3! = 6 • FirstPart+SecondPart+ThirdPart = 1*1*2*3 = 6 • The case for n = 3 is thus true.

  32. ForLoop Example - 4 • Next, use proof by induction to verify performance at all higher values of n. • Assume that operation is verified for n = k. • ForLoop = k! and • FirstPart+...LastPart = 1*1*2*3*4*...*k = k!

  33. ForLoop Example - 5 • Now, for k+1 • ForLoop = k!*(k+1) = (k+1)! • FirstPart+...LastPart = 1*1*2*3*4*...*k*(k+1) • = (k+1)! • The program is thus valid for k+1 so, by induction, it is valid for all values of n => 1.

  34. WhileLoop Verification - 1 • The WhileLoop is assumed to have the following form • WhileLoop • while WhileTest • begin • LoopPart • end

  35. WhileLoop Verification - 2 • Question 1. Is loop termination guaranteed for any argument of WhileTest? • Question 2. When WhileTest is true does • WhileLoop = LoopPart followed by WhileLoop? • Question 3. When WhileTest is false does • WhileLoop = identity?

  36. WhileLoop Example - 1 • Use the first part of a program to reduce a floating point number to standard form. • Exp is a positive integer • Mant is to be made a positive non- zero digit before the decimal point. • void NormalizeReal(double& Mant, int& Exp, int NegEx) • while (Mant >= 10) • { Mant = Mant/10; • if (NegEx) Exp=Exp-1; else Exp = Exp + 1; }

  37. WhileLoop Example - 2 • Since the value of Exp is not material, there are 3 cases of interest. • Each of these cases is examined for NegEx true and NegEx false. • 1. Mant <10 • 2. Mant = 10 • 3. Mant > 10

  38. WhileLoop Example - 3 • Question 1. Is loop termination guaranteed for any argument of WhileTest? • Case 1. Mant < 10, the while test fails immediately, so the loop terminates. • Case 2. Mant = 10, the while test will fail after one cycle, thus terminating the loop. • In both cases, this happens regardless of the value of NegEx.

  39. WhileLoop Example - 4 • Question 1, continued • Case 3. Mant > 10, can be shown by induction • Mant = k*10**n, where 1 <= k < 10 • when n = 1, the test fails after 1 cycle • if the test is true for some value n=m then after m cycles, Mant = k • thus for n=m+1, after m cycles, Mant = k*10 and the WhileTest will fail in one more cycle • Since the value of NegEx does not affect the result, question 1 is satisfied forall n => 1.

  40. A WhileLoop Example - 5 • Question 2. When WhileTest is true does • WhileLoop = LoopPart followed by WhileLoop? • Case 1. Mant < 10, is not pertinent since WhileTest is never true.

  41. A WhileLoop Example - 6 • To examine the other cases, check that this program gives the same result as WhileLoop (the check is described in the next foil). • Mant = Mant/10; • if (NegEx) Exp=Exp-1; • else Exp = Exp + 1; • while (Mant >= 10) • { Mant = Mant/10; • if (NegEx) Exp=Exp-1; • else Exp = Exp + 1;}

  42. A WhileLoop Example - 7 • Question 2. When WhileTest is true does • WhileLoop = LoopPart followed by WhileLoop? • For case 2. Mant = 10, the value of NegEx is immaterial • WhileLoop has one cycle, leaving Mant = 1 • LoopPart followed by WhileLoop starts with Mant = 10, LoopPart leaves value 1, and WhileLoop leaves the same value 1 • Since both give the same result, the answer is true.

  43. WhileLoop Example - 8 • Question 2 - Continued • Case 3. Mant > 10, can be shown by induction. • Mant = k*10**n, where 1<=k<10 • when n = 1, WhileLoop would take one cycle and the WhileLoop following the LoopPart would take none • when n > 1 the While Loop following LoopPart would take one less cycle with the same result • Question 2 is thus satisfied for all cases.

  44. WhileLoop Example - 9 • Question 3. When WhileTest is false does WhileLoop = identity? • Case 1. Mant < 10 is the only case where WhileTest is false. • Since, in this case the program is not executed, the conditions are left undisturbed which is the definition of identity. • Question 3 is thus satisfied.

  45. RepeatUntil Verification - 1 • RepeatUntil has the following form • RepeatUntil • begin • LoopPart • UntilTest • end

  46. RepeatUntil Verification - 2 • Repeat-until verification is satisfied if the answers to the following 3 questions are true. • Question 1. Is loop termination guaranteed for any argument of UntilTest? • Question 2. When the UntilTest after LoopPart is false, does • RepeatUntil = • LoopPart followed by RepeatUntil?

  47. RepeatUntil Verification - 3 • Question 3. When UntilTest after LoopPart is true, does • RepeatUntil = LoopPart? • The repeat-until verification follows the same logical format as WhileLoop verification so no example is given.

  48. Comments on Program Verification - 1 • If done correctly, these verification methods are quite general and can guarantee program correctness. • The verification methods are tricky and require skill. • While you should attempt to use these methods, when in doubt, check with a trace table or an execution table.

  49. Comments on Program Verification - 2 • With practice, you can do program verification quite rapidly. • The careful examination of all possible cases will quickly reveal most defects. • The loop termination test is critical because endless loops are hard to identify with other methods.

  50. Lecture #13 Assignment • Start work on the final report. • Update the process used to develop the mid-term report and use it to develop the final report. • Turn in the updated process, the plan for the work, the actual data for the work, and the final report. • Consult the report specifications in Appendix D.

More Related