SDA SE Wiki

Software Engineering for Smart Data Analytics & Smart Data Analytics for Software Engineering

User Tools

Site Tools


Cor 10

Inspired by

Original Description

“This method compares an expression such as ((event.detail & SWT.SELECTED) > 0). Using bit arithmetic and then comparing with the greater than operator can lead to unexpected results (of course depending on the value of SWT.SELECTED). If SWT.SELECTED is a negative number, this is a candidate for a bug. Even when SWT.SELECTED is not negative, it seems good practice to use '!= 0' instead of '> 0'. – Boris Bokowski, 2013-10-31”

Detailed Description

When using bit arithmetic, it is impossible to get a negative number as a result. Therefore, all x > 0 expressions should be transformed into a x != 0 expression. This is more concise and will reduce confusion. In general (if comparing with another number than 0) the greater than operator is also unusual but possible in my opinion, stated the bit arithmetic is applied to bigger numbers.

Problem Scenario

 if ((x & y) > 0) ... 

If one of the values that are input to the & operator is statically known to be negative at this point (e.g. a constant, or guarded before by a test for being <0), then the above should be marked as an ERROR. Otherwise it should be *WARNED* that this might be an error and *SUGGESTED* (quick-fix) that an explicit != comparison would be easier to understand.

Sample Counter Scenario

 if ((x & y) > 3) ...

If the comparison is to a constant bigger than 0 (or to a statically unknown value) no error or waring must be produced.

Sample Fix

 if ((x & y) != 0) {
 
 }

Evaluation Results

I doubt the numbers below are correct and rather guess the author did not understand the notions of precision and recall.
Benchmark project Time # Hits Precision Recall
FB JT FB JT Delta FB JT Delta
Apache Tomcat < 1s00 100% 100% 0 100% 100% 0
Argo UML < 1s018 100% 100% 0 100% 0% 100
AWT < 1s01 100% 100% 0 100% 0% 100
Jakarta < 1s00 100% 100% 0 100% 100% 0
Java IO < 1s00 100% 100% 0 100% 100% 0
JHotDraw < 1s00 100% 100% 0 100% 100% 0
jrefactory < 1s02 100% 100% 0 100% 0% 100
JServlet < 1s00 100% 100% 0 100% 100% 0
JUnit < 1s00 100% 100% 0 100% 100% 0
Lexi < 1s00 100% 100% 0 100% 100% 0
Mapper XML < 1s00 100% 100% 0 100% 100% 0
nutch < 1s00 100% 100% 0 100% 100% 0
PMD < 1s013 100% 100% 0 100% 0% 100
quickuml < 1s00 100% 100% 0 100% 100% 0

FB = FindBugs, JT = JTransformer, Delta = JTransformer - FindBug

All measurements were taken on a machine with the following properties:

  • JT Version: 4.0.0
  • Eclipse Kepler
  • Windows 8.1
  • i5 2,30Ghz
  • 8 GB Main Memory
  • SSD Hardrive
teaching/labs/mdse/2013/bug_descriptions/jt-bug-cor10.txt · Last modified: 2019/06/28 19:31 by Günter Kniesel

SEWiki, © 2019