[Soot-list] Bug in shimple transformation ?

[Jonas.Lundberg at msi.vxu.se]

Hi,

My name is Jonas Lundberg and I'm implementing a points-to 
analysis using Soot/Shimple as a front-end. I have had 
the same problem as Guillaume reported earlier on - missing 
phi-functions for certain multi-valued references.

My contribution is a way to identify cases where the analysis 
goes wrong.(I know, a solution should have been much better.) 
Anyway, in my analysis I find situations where the only possible 
target of a call a.m(), or a field access a.f is the null object 
(i.e. Pt(a) = {null}). This happens about 5-15 times for each 
program I try to analyze. I have looked at the Shimple 
code for a few methods where this happens, all of them have 
missing phi-functions.

A few examples
- java.io.BufferedReader.readLine(boolean)
  Several calls to r1 where r1 = null

- java.util.AbstractMap.remove(java.lang.Object) 
  A call r3.<java.util.Map$Entry: java.lang.Object getValue()>()   
where r3 = null
  
- java.text.SimpleDateFormat.compile(java.lang.String)
  Several calls to r3 where r3 = null
  
I hope that these examples can be of some help when testing 
the next "Nightly build" (which I hope will appear very soon).

Thanks in advance - Jonas
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.CS.McGill.CA/pipermail/soot-list/attachments/20051108/4a9d9812/attachment.htm