[Soot-list] Building a call graph: when?

[William Benton]

Hi all,

My code adds a whole-program transformation to the whole-shimple pack  
and a intraprocedural transformation to the shimple pack.  I also  
enable several of the Jimple optimizations in the jb pack.  Currently,  
the custom whole-program transformation builds a call graph (using  
spark), which is then used by the intraprocedural transformation.

I've noticed that many invocation statements aren't the source of any  
call graph edges when I examine each statement in the intraprocedural  
transformation.  Is this likely because the Jimple optimizations  
create new Units that weren't present when the call graph was  
constructed?  In general, what is the appropriate time to build a call  
graph?


thanks,
wb