CVS log for java/classes/org/w3c/util/ThreadCache.java

Request diff between arbitrary revisions

use notify instead of notifyall when one thread becomes available in the pool

Fixed import statements and removed extra empty lines

backport of 2.1 stability changes

updated synchronization

fixed threadcache to use a fifo, also some notify() calls were replaced by notifyAll() calls due to some broken implementations of notify().

fixed threadcache to use a looped-list, also some notify() calls were replaced by notifyAll() calls due to some broken implementations of notify().

Added WaitForCompletion(), it waits until all the threads have finished their
job (dead threads, or put in the freelist).

small optim to avoid too much thread creation near the overload zone

first commit for org.w3c

- Fixed a number of bugs...

- Removed traces (bug fixed ?)

- Fixed last bug (could write a book about it), removed traces.

- Previous bug fix in some other way (weird)

- Fixed a race condition:
  The problem is quite tricky: upon creating a new thread for the
  cache, the code would do:
      thread = new Thread();
      thread.start();
  Starting the newly created thread, *before* it was assigned an
  object to run on (through CacheThread::wakeup). If the started
  thread reach the runner check before wakeup was called, it would
  termintae itself (under heavy load), but still leaving the cache
  under the impression it would run for the given object !
  The fix is to make sure, when a thread is started, that it will run
  *at least* for one runnable (through the firsttime flag)

- Added a setGrowAsNeeded method

- Fixed a bug: when a thread is dead it has to notify any other thread
  waiting for a free thread.
- Added documentation.

Some new features, and some debugging

Enhancements, and bug fixes

- Added ThreadCache (a simple ThreadCache)
- Added Unix system calls