Paul,

The change sets seem to be completely broken. There are tons of exclamation
marks at weird parts in the files so at least I can't file them in
successfully. Could you please repost the stuff?!

Thanks,
  - Andreas

> -----Original Message-----
> From: Paul McDonough [mailto:wnchips at yahoo.com]
> Sent: Thursday, August 17, 2000 3:45 PM
> To: squeak at cs.uiuc.edu
> Cc: recipient list not shown
> Subject: [FIX] exceptional joy ...
> 
> 
> Fellow Squeakers,
> 
> Please find attached two change sets which, taken
> together, fix all known bugs in the exception handling
> system (including ensured execution).  As a side
> effect, everyone's favorite method
> (BlockContext>>valueUninterruptably) goes away - so if
> you're particularly sentimental about that method, go
> have one last read-through before filing this in.
> 
> Anyway.  The first change set is one that Stephen Pair
> sent around a few months ago, modifying the
> Interpreter to support ensured execution.  Using a vm
> that (I think) Andreas Raab built, I have verified
> that it does indeed work as expected.
> 
> The second change set contains the actual
> modifications to the EHS.  In aggregate, it results in
> a slight reduction in the code base.
> 
> In the past, there has been some concern that the
> Interpreter modifications will slow down the vm.  I
> have not run any benchmarks, but I expect that the hit
> will be trivial:  if you look at Stephen's changes,
> you will observe that the additional overhead consists
> of a single equality test in
> Interpreter>>returnValue:to:, in the normal case. 
> _Only_ in the case of a nonlocal return-in-progress is
> there any significant additional processing (and prior
> to the modifications it didn't work anyway).  So imho,
> SqC should consider moving these modifications
> straight into the core vm stream.
> 
> Although I get to have the pleasure of announcing the
> results, I should observe that the Squeak EHS has been
> quite a collaborative effort.  Exceptions are a nasty
> complicated hairy beast of a subset of Smalltalk, and
> without the work done by the Richard Harmon and the
> Camp Smalltalk ANSI group we wouldn't have had all the
> unit tests that give me the confidence that this thing
> actually works - thanks guys!  Allen Wirfs-Brock and
> the other authors of the ANSI document get credit for
> giving us a target.  Nice semantics!
> 
> Stephen Pair and Craig Latta contributed code to the
> effort, and in addition both of them helped me to
> think through some of the more twisted questions. 
> Dan, Andreas, and David Caster also threw valuable
> insights into the mix, and Steve Rees confirmed the
> existence of the very-very-evil 'nested ensure: bug'.
> 
> Finally, thanks to exobox for putting on the finishing
> touches, the results of which are in the
> aforementioned second change set.  I hope I'm not
> leaving anyone out, but if so, apologies!
> 
> Well, I forget myself, this was not meant to be an
> Oscar acceptance speech, but merely an announcement to
> accompany some code.  Here 'tis.
> 
> Paul
> 
> __________________________________________________
> Do You Yahoo!?
> Send instant messages & get email alerts with Yahoo! Messenger.
> http://im.yahoo.com/
>