[cpp-threads] Re: Berlin meeting

[Lawrence.Crowl at Sun.com]

"Boehm, Hans" <hans.boehm at hp.com> writes:
 >Thanks.  Will you turn this into a committee paper, or post it somewhere?

Already done.  It will be in the post-Berlin mailing.

 >As I recall, we also spent some time discussing arguments for special syntax
 >for various constructs.  Some of that is embedded in your minutes.  But I
 >didn't see all of it.  From what I remember, the main points were:
 >
 >Fork: Argument marshalling is an issue with a pthread like interface which
 >doesn't have syntax support.  Anything that relies more on objects (or the
 >proposed C++ lambdas) has C compatibility issues.
 >
 >Join: Can be viewed as a library call, but there needs to be magic behind
 >the implementation to propagate exceptions, since we don't know the
 >exception type.  It cannot be implemented as purely a library.
 >
 >mutex lock/unlock: Java-like syntax is more convenient in a C setting.  In
 >either a C or C++ setting, it probably makes condition variable usage more
 >convenient, since the lock can be implicit.  But you need library versions
 >anyway to deal with some more obscure cases.

This is consistent with my memory as well.  The join exception issue
was mentioned briefly in the minutes, but the rest of the discussions
occurred before I started taking minutes.

 >By the time I left, it didn't look to me like this decisively resolved the
 >question of whether any or all of these could be exposed purely as library
 >calls.

It was not resolved at Berlin.

  Lawrence Crowl             650-786-6146   Sun Microsystems, Inc.
                   Lawrence.Crowl at Sun.com   16 Network Circle, UMPK16-303
           http://www.Crowl.org/Lawrence/   Menlo Park, California, 94025