Re: Data flow analyzer for Java bytecode

Paolo Maggi (p.maggi@studenti.to.it) wrote:
: I'm attending an Italian University and, for my graduation thesis, I
: need to design and develop a inter- and intra-procedural data flow
: analyzer for Java bytecode.


I'm interested in knowing what you hope to gain by making the analysis inter-procedural. The verifier described in the VMspec, implemented in Sun's JDK as well as (most probably) the kimera verifier all treat each method independently. The reasons for this are, I think,
- because classes are managed, loaded and linked separately
- it is not necessary because all needed information is available to each
methods (i.e. signatures of other methods invoked and fields accessed)


Perhaps a slightly less restrictive verifier could be arrived at, but such
flexibility could not be used in practice because other verifiers (like Sun's)
wouldn't allow it. And even then, I wonder what flexibility would be gained.


: Do you know where I could find some information about that?


: I'd also like to understand the way of constructing the control flow
: graph of a method when it contains jsr-ret blocks. Thank you very
: much for your help.


In my opinion the VM spec is quite poor on the subject of verification. You
kind of have to get Sun's source code. Much more information can be extracted
from that, but it's a difficult process.


If you like formal stuff:
http://www.research.digital.com/SRC/personal/
    Martin_Abadi/Papers/jvm-short-preprint.ps
http://Theory.Stanford.EDU/~freunds/jvm-full.ps
http://www.informatik.uni-bremen.de/~qian/abs-fsjvm.html


Good luck
--


Stéphane Doyon
B.Sc informatique
s.doyon@videotron.ca
--

Return to the comp.compilers page.
Search the comp.compilers archives again.