RE: JavaMemoryModel: Proposed wording on lost interrupts

From: David Holmes (dholmes@dltech.com.au)
Date: Wed Aug 06 2003 - 19:43:55 EDT


> Behavior (a) is the preferred behavior, and the strong
> version of the
> wait/interrupt semantics allows only behavior (a). Weaker
> versions of
> the wait/interrupt semantics would allow either behaviors (a) and
> (b), or even any of the three.

There are some secondary issues concerning how an implementation that
would allow c) causes itself to act like b) in the above scenario.
This is the case where a notification must be reissued. Depending on
how the notification is reissued you can get some non-obvious
interactions when considering scenarios more complex than the basic
one given.

David Holmes

-------------------------------
JavaMemoryModel mailing list - http://www.cs.umd.edu/~pugh/java/memoryModel



This archive was generated by hypermail 2b29 : Thu Oct 13 2005 - 07:00:50 EDT