hmm... looking at the current svn HEAD code base, we definitely need to
clean up how Thread.join works.  From a top-level, keeping part of Java
thread state in java objects and part in C structs is confusing and likely
to contain bugs.  I favor moving all thread state into C structs and writing
Java native methods that support java.lang.Thread.  And also substantially
simplifying the lock protocol on the info contained in the thread data
structs.

Given the above, does it make sense to apply a patch that goes exactly the
opposite direction?  That is, removes dead native join implementation.  I
would hate to have to reverse directions and add Thread.join() code back in
a few months.

--
Weldon Washburn
Intel Enterprise Solutions Software Division

Reply via email to