RE: [abc] oopsla outline

From: Oege de Moor <Oege.de.Moor@comlab.ox.ac.uk>
Date: Thu Mar 02 2006 - 15:58:26 GMT

Hi Eric,

Following on from an abc meeting we just had on this end,
we were hoping you could chase up the following items
for the paper:

* provide scripts for j-lo to process jhotdraw+safeenum

   Pavel says he couldn't get jhotdraw to terminate
   when instrumented with j-lo. you mentioned you made
   some improvements, so perhaps you could try again?
   If necessary, recommend a number of iterations that
   can be processed by j-lo, so we can use that number
   for the other versions.

   Does your version produce
   memory measurements in the same way as the other versions do?
   Generally, it would be great if the j-lo version was
   ready for comparison both in time and space.

* get hawk to process jhotdraw+safeenum? I guess our
   only hope is Klaus willing to help out.

* hashcode tm on approve:
   does abc now properly process approve? You mentioned
   earlier that you made it skip some problematic class.
   If that is still so, have you checked that class does
   not contain any hashing-related stuff?

* lock order reversal.
   Nobody on this end understands the tracematch you wrote -
   you've thought about this problem before, but we haven't.
   Can you send a careful explanation to the list, starting
   with the problem, a few examples where it should match
   and shouldn't, and so on. Sorry to be slow!

   Can you post the results of running it, so we can all see
   both time and memory behaviour? [we especially want to
   know about the second column :-)]

   Is this problem expressible in PQL? It would be really
   good if we could run a comparison!

In the meantime, Julian and Pavel are working round the clock
to get indexing finished.

Ganesh is implementing thisJoinPointIdentity
for use in the Jigsaw locknunlock-in-same-method
variations, to determine whether a backend optimisation
of thisJoinPoint would be worthwhile.

Neil is in charge of the "third table" (on the different
ways of expressing CFL properties, via thisJP+cflow, via
cflowdepth, and lots of other ways. He is also preparing
the AOSD talk on open modules.

Many thanks!

-Oege
Received on Thu Mar 02 15:58:44 2006

This archive was generated by hypermail 2.1.8 : Tue Mar 06 2007 - 16:13:27 GMT