Effective and Efficient Localization of Multiple Faults using Value Replacement Dennis Jeffrey

Neelam Gupta

Rajiv Gupta

[email protected]

[email protected]

[email protected]

Presented by Dennis Jeffrey Department of Computer Science and Engineering The University of California, Riverside

25th IEEE International Conference on Software Maintenance Edmonton, Alberta, Canada September 24, 2009 Effective and Efficient Localization of Multiple Faults using Value Replacement

Value Replacement: Overview Dynamic state alteration technique to locate faulty program statements [Jeffrey et. al., ISSTA 2008] INPUT: Faulty program and test suite (1+ failing runs)

TASK: (1) Perform value replacements in failing runs (2) Rank program statements according to collected information

OUTPUT: Ranked list of program statements Effective and Efficient Localization of Multiple Faults using Value Replacement

2 / 26

Alter State byy Replacing g Values Passing Execution

Correct Output Effective and Efficient Localization of Multiple Faults using Value Replacement

3 / 26

Alter State byy Replacing g Values Failing Execution

ERROR

Incorrect Output Effective and Efficient Localization of Multiple Faults using Value Replacement

4 / 26

Alter State byy Replacing g Values Failing Execution: Altered State

ERROR Statement: S Instance: I Original Set of Values: ORIG Alternate Set of Values: ALT

REPLACE VALUES

IVMP: If correct: Correct? / Incorrect?

Stmt S, S Instance I: ORIG Æ ALT

Effective and Efficient Localization of Multiple Faults using Value Replacement

5 / 26

Value Replacement: Details Replace values at different statement instances in failing runs to search for IVMPs Statements with an IVMP in more failing runs are more likely to be faulty Alternate value sets taken from profiling info

Rank program statements in decreasing o de o c ous ess order of susp suspiciousness Suspiciousness: the # of failing runs in which the given statement is associated with an IVMP Effective and Efficient Localization of Multiple Faults using Value Replacement

6 / 26

Value Replacement: Example 1: read (x, y); 2: a := : x + y; 3: if (x < y) 4: write (a); else 5: write (a + 1);

Test Case (x, y) [A]

(1, 1)

3

1

[B]

(0, 1)

1

-1

( 1 0) [C] (-1,

-1 1

-1 1

(0, 0)

1

1

[D] Test Case [A] IVMPs:

Test Case [B] IVMPs:

stmt 1, inst 1: ( {x=1, y=1}Æ{x=0, y=1} ) stmt 1, inst 1: ( {x=1, y=1}Æ{x=0, y=0} ) stmt t t2 2, iinstt 1 1: ( {{x=1, 1 y=1, 1 a=2}Æ{x=0, 2}Æ{ 0 y=1, 1 a=1} 1} ) stmt 2, inst 1: ( {x=1, y=1, a=2}Æ{x=0, y=0, a=0} ) stmt 5, inst 1: ( {a=2, output=3}Æ{a=0, output=1} )

stmt 1, inst 1: ( {x=0, y=1}Æ{x=-1, y=0} ) stmt 2, inst 1: ( {x=0, y=1,a=1}Æ{x=-1, y=0,a=-1} ) stmt t t 4, 4 iinstt 1 1: ( {{a=1, 1 output=1}Æ{a=-1, t t 1}Æ{ 1 output=-1} t t 1} )

stmts t t with ith IVMPs: IVMP {1, {1 2, 2 4}

stmts with i h IVMPs: IVMP {1, {1 2, 2 5}}

MOST LIKELY TO BE FAULTY

{1 2} {1,

Actual Output Expected Output

{4 {4, 5}

{3}

LEAST LIKELY TO BE FAULTY

Effective and Efficient Localization of Multiple Faults using Value Replacement

7 / 26

Value Replacement: Results Highly Effective Precisely locates 39 / 129 errors (30.2%) Most effective previously known: 5 / 129 (3 (3.9%) 9%)

Limitations Assumes multiple p failing g runs are caused by y the same error Can require significant computation time to search for IVMPs Effective and Efficient Localization of Multiple Faults using Value Replacement

8 / 26

Handling g Multiple Errors: Goals Effectively locate multiple simultaneous errors Iteratively compute a ranked list of statements to find and fix one error at a time Three variations of this technique MIN FULL PARTIAL

Effi i l search h ffor IVMPs IVMP Efficiently Improve efficiency without impacting effectiveness Two motivating T ti ti observations b ti Redundancy Independence Effective and Efficient Localization of Multiple Faults using Value Replacement

9 / 26

Multiple-Error Techniques Single Error

Multiple Errors (MIN)

Faulty Program and Test Suite

Faulty Program and Test Suite

Value Replacement

Value Replacement

Ranked List of Program Statements

Ranked List of Program Statements

Developer Find/Fix Error

Done

Developer Find/Fix Error

Yes

Failing Run Remains?

Effective and Efficient Localization of Multiple Faults using Value Replacement

No

Done

10 / 26

Multitple-Error Techniques Multiple Errors (FULL)

Multiple Errors (PARTIAL)

Faulty Program and Test Suite

Faulty Program and Test Suite

Partial Value Replacement

Value Replacement Ranked List of Program Statements

Ranked List of Program Statements

Developer Find/Fix Error

Yes

Failing Run Remains?

Developer Find/Fix Error

No

Done

Yes

Failing Run Remains?

Effective and Efficient Localization of Multiple Faults using Value Replacement

No

Done

11 / 26

PARTIAL Technique Step 1: Initialize ranked lists and locate first error For each statement s, compute a ranked list by considering only failing runs exercising s Report ranked list with highest suspiciousness value at the front of the list

Step 2: Iteratively revise ranked lists and locate each remaining g error For each remaining failing run that exercises the statement just fixed, recompute IVMPs Update any affected ranked lists Report ranked list with the most different elements at list compared to previously-selected previously selected lists the front of the list, Effective and Efficient Localization of Multiple Faults using Value Replacement

12 / 26

PARTIAL Technique: Example Program (2 faulty statements)

Failing Run

Execution T Trace

Statements with i h IVMPs IVMP

A

(1, 2, 3, 5)

{2, 5}

B

(1, 2, 3, 5)

{1, 2}

C

(1, 2, 4, 5)

{2, 4, 5}

1

2 C Computed t dR Ranked k d Li Lists: t ((statement t t tsuspiciousness) 3

4

5

Report list 1, 2, or 5 (assume 1) Î Fix faulty statement 2

1

23, 52, 11, 41, 30

[based on runs A, B, C]

2

23, 52, 11, 41, 30

[based on runs A, B, C]

3

22, 11, 51, 30, 40

[based on runs A, B]

4

21, 41, 51, 10, 30

[based on run C]

5

23, 52, 11, 41, 30

[based on runs A, B, C]

Effective and Efficient Localization of Multiple Faults using Value Replacement

13 / 26

PARTIAL Technique: Example Program (1 faulty statement) 1

Failing Run

Execution T Trace

Statements with i h IVMPs IVMP

C

(1, 2, 4, 5)

{4}

Computed Ranked Lists: (statementsuspiciousness) 2 3

4

5

2

22, 11, 41, 51, 30

[based on runs A, B, C] (C updated)

3

22, 11, 51, 30, 40

[based on runs A, B]

4

41, 10, 20, 30, 50

[based on run C]

5

22, 11, 41, 51, 30

[based on runs A, B, C] (C updated)

(no updates) (C updated)

Report list 4 Î Fix faulty statement 4 Î Done Effective and Efficient Localization of Multiple Faults using Value Replacement

14 / 26

Improving g Efficiency y of IVMP Search Original Execution

Regular Value Replacement Executions (x 6)

stmt instance 1

(x 4)

stmt instance 2

(x 2)

stmt instance 3

(assume 2 alternate value sets at each stmt instance)

(value replacements are independent of each other) (portions of original execution are duplicated multiple times)

Efficiency Improvements: (1) Fork child process to do each value replacement in original failing execution (2) Perform value replacements in parallel

Effective and Efficient Localization of Multiple Faults using Value Replacement

15 / 26

Improving g Efficiency y of IVMP Search With Redundant Execution Removed

( d (no duplication li ti off any portion ti off original execution)

With Parallelization

(t t l ti (total time required i d to t perform f all ll value l replacements is reduced)

Effective and Efficient Localization of Multiple Faults using Value Replacement

16 / 26

Experimental Evaluation Original Siemens Benchmark Programs

Program

LOC

# Faulty Ver.

Test Case Pool Size

t tcas

138

41

1608

totinfo

346

23

1052

sched

299

9

2650

sched2

297

9

4130

ptok

402

7

4130

ptok2

483

9

4115

replace

516

31

5542

Effective and Efficient Localization of Multiple Faults using Value Replacement

17 / 26

Experimental Evaluation Experimental Benchmark Programs

Program

# 5-Error Faulty Versions

Average Suite Size (# Failing Runs / # Passing Runs)

tcas

20

11 (5 / 6)

totinfo

20

22 (10 / 12)

sched

20

29 (10 / 19)

sched2

20

30 (9 / 21)

ptok

2

32 (8 / 24)

ptok2

11

29 (5 / 24)

replace

20

38 (9 / 29)

Each faulty program contains 5 seeded errors, each in a different stmt Each faulty y program p g is associated with a stmt-coverage g adequate q test suite such that at least one failing run exercises each error Effective and Efficient Localization of Multiple Faults using Value Replacement

18 / 26

Techniques Compared (MIN) Only compute ranked list once (FULL) Fully recompute ranked list each time (PARTIAL) Compute IVMPs for subset of failing runs and revise ranked lists each time (ISOLATED) Locate each error in isolation

Effective and Efficient Localization of Multiple Faults using Value Replacement

19 / 26

Metric for Comparison Score for each ranked statement list rank of the faulty stmt

size of list

x 100%

size of list

Represents percentage of statements that need not be examined before error is located Higher score is better High Hi h Suspiciousness

Low L Suspiciousness

High Hi h Suspiciousness

Low L Suspiciousness

Higher Score Effective and Efficient Localization of Multiple Faults using Value Replacement

Lower Score 20 / 26

Effectiveness Results 90 85 80

Isolated Full Partial Min

75 70 65 60

replace

ptok2

ptok

sched2

sched

50

totinfo

55 tcas

Avg. Scorre per Ranked Lis A st (%)

Effectiveness Comparison p of Value Replacement p Techniques q

Effective and Efficient Localization of Multiple Faults using Value Replacement

21 / 26

Efficiencyy Results Time to Search for IVMPs for Each Faulty y Program g

500

Multiple errors can be located in minutes.

400 300

Full Partial Min

Previously, single errors could be located in hours.

200

replace

ptok2

ptok

sched2

sched

0

totinfo

100 tcas

Avg.. Time (se econds)

600

Effective and Efficient Localization of Multiple Faults using Value Replacement

22 / 26

Prior Techniques for Locating g Errors Program slicing Pruning Dynamic Slices with Confidence [Zhang [Zh et. t al.l PLDI 2006] Failure-Inducing Chops [Gupta et. al. ASE 2005]

Invariant-based techniques Daikon [Ernst et. al. IEEE TSE Feb. 2001] AccMon [Zhou et. al. HPCA 2007]

Effective and Efficient Localization of Multiple Faults using Value Replacement

23 / 26

Prior Techniques for Locating g Errors Statistical techniques Cooperative Bug Isolation [Ben [B Liblit d doctoral t l di dissertation, t ti 2005] SOBER [Liu et. al. FSE 2005] Tarantula [Jones et. al. ICSE 2002] COMPUTE

RESULT

State-alteration State alteration techniques

RESULT

Effective and Efficient Localization of Multiple Faults using Value Replacement

24 / 26

Prior State Alteration Techniques Delta Debugging [Zeller et al. FSE 2002, TSE 2002, ICSE 2005] Search in space for values relevant to a failure Search in time for failure cause transitions

Predicate Switching [Zhang et. al. ICSE 2006] Alt predicate Alter di t outcome t tto correctt failing f ili output t t

E ti Suppression S i [Jeffrey et. al. ICSM 2008] Execution Suppress effects of memory corruption during execution to locate root causes of memory errors

Effective and Efficient Localization of Multiple Faults using Value Replacement

25 / 26

Conclusions 3 technique variants to locate multiple errors Minimal computation (more efficient, efficient less effective) Full recomputation (more effective, less efficient) Partial recomputation (more balanced effectiveness/efficiency)

2 techniques to improve efficiency of IVMP search Remove redundant R d d t program execution ti Parallelize the search

Multiple simultaneous errors can be effectively located in minutes in the benchmark programs. Previously, single errors could be effectively located in hours using the same benchmarks.

Effective and Efficient Localization of Multiple Faults using Value Replacement

26 / 26

Effective and Efficient Localization of Effective and ...

Delta Debugging[Zeller et al. FSE 2002, TSE 2002, ICSE 2005]. Search in space for values relevant to a failure. Search in time for failure cause transitions.

280KB Sizes 1 Downloads 194 Views

Recommend Documents

Efficient and Effective Video Copy Detection Based on Spatiotemporal ...
the Internet, can be easily duplicated, edited, and redis- tributed. From the view of content ... in this paper, a novel method for video copy detection is proposed. The major ...... images," IEEE International Conference on Computer. Vision, 2005.

Efficient and Effective Similarity Search over Probabilistic Data ...
To define Earth Mover's Distance, a metric distance dij on object domain D must be provided ...... Management of probabilistic data: foundations and challenges.

Effective and Efficient Fingerprint Image Postprocessing
Email: [email protected] [email protected] [email protected] Abstract. Minutiae extraction is a crucial step in an automatic fingerprint identification system.

Efficient and Effective Similarity Search over Probabilistic Data Based ...
networks have created a deluge of probabilistic data. While similarity search is an important tool to support the manipulation of probabilistic data, it raises new.

Efficient and Effective Video Copy Detection Based on ...
Digital videos, which have become ubiquitous over ... The merit of the ordinal signature is the robust- ... butions, while the centroid-based signature computes.

What's on TV Tonight? An Efficient and Effective ...
Dec 30, 2008 - approach cannot deal with data sparsity in a principled manner. ..... case of recommender systems, is that it can provide the best low-rank ..... living room by using their PDA, cellular phone or set-top-box with internet access.

Effective, Efficient, Fair - Texas Public Policy Foundation
May 21, 2004 - Effective, Efficient, Fair: Paying For Public Education In Texas. Texas Public Policy ..... teacher pay, adjusted for fringe benefits and vacation periods, is typical of comparably educated professionals. 12 ..... Available online at:.

The Relevance of Effective Database System for Efficient Office ... - IJRIT
Keyword: Database Security, Database System, Office Management, Oracle 11g, SQL, Tertiary Institution .... maintenance and problems diagnostic functionality, provide automatic backup, secure storage facilities, provide data ... RAID (Redundant Array