Skip to Content.
Sympa Menu

svadev - Re: [svadev] DSA and complete programs

svadev AT lists.siebelschool.illinois.edu

Subject: Svadev mailing list

List archive

Re: [svadev] DSA and complete programs


Chronological Thread 
  • From: Will Dietz <willdtz AT gmail.com>
  • To: "Ryan M. Lefever" <lefever AT illinois.edu>
  • Cc: svadev AT cs.uiuc.edu
  • Subject: Re: [svadev] DSA and complete programs
  • Date: Mon, 11 Oct 2010 23:01:30 -0500
  • List-archive: <http://lists.cs.uiuc.edu/pipermail/svadev>
  • List-id: <svadev.cs.uiuc.edu>

On Mon, Oct 11, 2010 at 10:34 PM, Ryan M. Lefever
<lefever AT illinois.edu>
wrote:
> If you have a full program, i.e., there are no external call sites.  Am
> I correct that after running TD DSA, all nodes should be marked
> complete?  It seams to me that marking a node complete is a function of
> reachability.  In the case where no functions are external, you have
> full reachability information by propagating DSA info from callees to
> callers (BU) and then from callers to callees (TD, which requires BU).
>
> Regards,
> Ryan
> _______________________________________________
> svadev mailing list
> svadev AT cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/svadev
>

In short, TD on 2.6 has a bug where it marks *everything* as
incomplete. I can ship you a patch that fixes it, but it the
resulting analysis will not handle var-args codes correctly (among
whatever else may or may not be wrong in that release, I cannot speak
authoritatively on the correctness of DSA 2.6, beyond this
completeness bug).

As for what version people should be using, and what we're maintaining
or officially supporting...I'll have to defer to the rest of the group
(Vikram? John?) on that, I'm new here :).

I'll work on the patch tomorrow (your the second person looking for
such a patch), as I'm a bit swamped tonight :).

Thanks for the detailed report, and have a good one.

~Will





Archive powered by MHonArc 2.6.16.

Top of Page