jikesrvm-core-bounces@lists.sourceforge.net wrote on 06/25/2007 01:19:35 PM:

> <snip>

> various ways to report violation
> </snip>
all ok with me.  I'd bias towards making it a more fatal error by default, but I know sometimes that can be annoying for debugging.

> 5) "fix" the compilers so that we don't generate warnings. For example,
> provide uninterruptible versions of new. There's probably a limit of how
> much of this can happen.

I think we'd actually have to do real work to enable any more cases and all the low-hanging fruit has been done already.  I suspect there's very little additional cases we could cover.  Uninterruptible is a very strong property: it is effectively an atomic section from the view of M-N threading and what happens on a single virtual processor.