Re: [abc] Sunday and next week

On Fri, Sep 03, 2004 at 10:47:45AM -0400, Prof. Laurie HENDREN wrote:
> Aske and I can work on the architecture part, then we will figure out
> who will actually present it - I don't mind if Aske does it because
> he was the guy thinking about the real architecture at first .... remember
> the white board in his office!
>
> Aske, can you think about this and also the outline of the CC paper. If
> you send me something by Monday before I leave (about 5pm your time on
> Monday), then I can grab it on my laptop and take it with me to the
> airport.
>
> Also, I can work with Ondrej on the Soot presentation part ... I would
> like to include a bit about why it is helpful to be working with Jimple
> and not bytecode, so if anyone has some ideas there, let us know.

I think we can use a subset of the PLDI 2003 tutorial slides for most of
it; I'll have the source for those with me. Unfortunately, two weeks
ago, when we moved, I recycled the 60 extra copies of the handouts that
we had left over from CASCON. We'll pick out a subset (unless we want a
3-hour talk on Soot) and figure out what to add.

In terms of examples illustrating the value of Jimple, one I can think
of is the thing Ganesh discovered about the ?: operator bug. Because
ajc operates on bytecode, it considers exactly one of foo()?bar():2 and
foo()?2:bar() to be constant. But I'm sure there are other examples. I
suppose it would be easier to see what Jimple makes easy compared to
bytecode if the bulk of our experience was with bytecode rather than
Jimple, and we knew what things that are hard on bytecode. It may be
worthwhile to look at ajc for the really hairy bits.

Ondrej

> Other things we might consider briefly discussing while at IBM - a
> time for a BOF at OOPSLA for ajc/abc stuff. Possible future
> interactions with IBM (UK and Canada).
>
> Cheers, Laurie
>
>
> >
> > Ondrej and Aske are arriving Saturday.
> >
> > On Sunday the plan is to get together at 11am at the Worcester lodge.
> > We'll wander over to Magdalen, to go punting. I'll stock up on bbq
> > materials etc. Please let me know whether you intend to join, so I
> > have an idea about numbers.
> >
> > We'll go to IBM on Thursday. We've rented a van, which Ganesh will
> > drive.
> >
> > There are the following things to do during the week:
> > a) prepare for the meeting with ajc team. We'll need
> > * some short talks:
> > aims of abc (me), architecture (Laurie?), frontend & polyglot (me),
> > Soot (Ondrej), advice weaver (Ganesh), around advice (Sascha).
> > * list of topics for discussion
> > - tricky points in semantics
> > - bugs in test suite
> > - future changes to aspectj
> > Please think about these matters now. It will be best if everyone
> > has their own list ready.
> >
> > b) start writing the AOSD paper about extensions - using Julian's
> > examples. This will be a high-level "howto use abc in your research"
> > paper. I hope we'll have a rough draft by Friday.
> >
> > c) start on the CC paper - on the architecture of abc.
> > Aske, you were going to sketch this?
> >
> > d) implement the cflow eliminator using Ondrej's analysis
> >
> > It will be a busy but exciting week!
> >
> > -Oege
> >
> >
> >
> >
> >
> >
>
>
Received on Fri Sep 3 20:41:31 2004