|CFG construction for Java bytecode methods firstname.lastname@example.org (1999-06-06)|
|Re: CFG construction for Java bytecode methods email@example.com (David Chase) (1999-06-12)|
|From:||David Chase <firstname.lastname@example.org>|
|Date:||12 Jun 1999 21:24:02 -0400|
Guy Laden wrote:
> Can one use the fact that a Java bytecode method has passed verification
> to simplify the computation of _precise_ 'ret' instruction targets,
> avoiding a flow analysis?
I'm not sure that you can. First of all, you aren't avoiding
flow analysis, since verification requires flow analysis.
Second, sometimes jsr and ret are used in quirky ways that
nonetheless pass flow analysis; I think you can sometimes
get different jsr's paired with a single ret, for instance,
or a ret that pops you out two levels (e.g., jsr saving return
address in a, followed by jsr saving return address in b,
followed by a "ret a" exiting both "blocks").
I know that we (not me personally) had a heck of a time
getting it right.
David Chase -- email@example.com
NaturalBridge LLC -- http://www.naturalbridge.com
Return to the
Search the comp.compilers archives again.