[Soot-list] Beginner's points-to question

[Chris Pickett]

Oh yeah, I remember now, thanks.  Knowing that we have a points-to set 
of size one where the allocation site is only run once is actually 
exploited in the MHP analysis in Soot, although not for must alias reasons.

Chris

Eric Bodden wrote:
> Nope, that's wrong. Size 1 means there is a unique creation site but
> unless you know that this site is visited once that does not tell you
> anything.
> 
> Eric
> 
> On 30/01/2008, Chris Pickett <chris.pickett at mail.mcgill.ca> wrote:
>> If both sets have size one then the variables must alias, surely?  I
>> think this would need an explicit representation of null in your sets.
>>
>> Chris
>>
>> Eric Bodden wrote:
>>> Hi, Richard. There is no whole-program must-alias analysis in Soot.
>>> (and I think there is none on this planet) Points-to sets can only be
>>> used to decide may-aliasing. Two variables cannot alias if their
>>> points-to sets do not intersect.
>>>
>>> Our SVN version of Soot has however an intraprocedural must-alias analysis:
>>>
>>> soot.jimple.toolkits.pointer.LocalMustAliasAnalysis
>>>
>>> Maybe that one is good enough for you?
>>>
>>> If you find an interprocedural must-alias analysis or wish to develop
>>> one, please let me know!
>>>
>>> Eric
>>>
>>> On 30/01/2008, Richard Warburton <richard.warburton at gmail.com> wrote:
>>>> Having been through the section of the brics guide on program analysis
>>>> on points-to analysis, I have to admit to a certain level of
>>>> confusion.    Primarily as to how to compute a must-aliasing relation
>>>> between two variables.  I am using SPARK.  If I have the PointsToSet
>>>> instances for each variable, I can check their possible intersection,
>>>> but how do I check their definite intersection?  If the phrasing of
>>>> this question seems to imply that I have misunderstood the points-to
>>>> implementation within soot, it would be quite helpful to see an
>>>> example of how to compute the may and must aliasing relations from
>>>> this analysis.
>>>>
>>>>   Richard Warburton
>>>> _______________________________________________
>>>> Soot-list mailing list
>>>> Soot-list at sable.mcgill.ca
>>>> http://mailman.cs.mcgill.ca/mailman/listinfo/soot-list
>>>>
>>>
> 
>