Re: Soot typing failure

[Etienne Gagnon <gagnon.etienne_m@uqam.ca>]

Hmmm.

I was wrong I think.  See below.

Etienne Gagnon wrote:
> The constraint checker "Is Right(TM)".  Your problem is probably that 
> typing is applied after some optimizations.  IT SHOULD NOT.  I DISAGREE with 
> the proposed patch that Ondrej happily applied.

I am not so sure anymore.  After checking my assumptions in
ConstrtaintCollector, I have to admit that the patch makes sense.


> SOOT PROBLEM #1
> ===============
>
>         PackManager.v().getTransform( "jb.a" ).apply( b );
>         PackManager.v().getTransform( "jb.ule" ).apply( b );
>
> is applied BEFORE typing in soot-2.0.1/soot/src/soot/JimpleBodyPack.java


Hmmm.  After reading the manual about these, they seem not to be destructing
valuable info for typing.  We can leave them as-is.  But, remember that typing
depends on having enough local variables to be able to assign different types
to different locals.  Merging locals too early can result in type inference
failures.


> SOOT PROBLEM #2
> ===============
>
> In the code:
>...
> of soot-2.0.1/soot/src/soot/JimpleBodyPack.java,  verification of typing
> failure is done EVEN WHEN typing is disabled!  (e.g. -p jb.tr
> enabled:false )


This still looks as a problem to me.

Etienne

--
Etienne M. Gagnon, Ph.D.             http://www.info.uqam.ca/~egagnon/
SableVM:                                       http://www.sablevm.org/
SableCC:                                       http://www.sablecc.org/