Name: nt126004 Date: 09/11/2002


FULL PRODUCT VERSION :
java version "1.4.0_01"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.4.0_01-b03)
Java HotSpot(TM) Client VM (build 1.4.0_01-b03, mixed mode)


FULL OPERATING SYSTEM VERSION :
glibc-2.2.2-10
Linux twright 2.4.2-2 #1 Sun Apr 8 20:41:30 EDT 2001 i686
unknown

A DESCRIPTION OF THE PROBLEM :
I'm using a Thread pool in my application. I'd like to
set the thread name for each runnable that I schedule out of
my pool. I'd also like thread dumps to show these names.

OptimizeIt reveals an oddity in "java.lang.Thread":
  a) Names are kept as "char[]"s instead of Strings
  b) Every call to "setName(String)" results in a clone of
     the passed String's character array.
  c) Every call to "getName()" results in yet another
     clone of the character array.
Lastly, "setName(String)" and "getName()" are final, so I
can't fix the problem by subclassing.

I see no need for this GC + performance overhead.


STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
See the test below.

EXPECTED VERSUS ACTUAL BEHAVIOR :
Calls to "setName(String)" that don't change the name should
be no-ops. Even if the name is being changed, it shouldn't
result in char[] / String allocations. Thread should keep a
pointer to the String, not the char[].

Calls to "getName()" should be cheap. Strings are immutable!

Serialization compatibilty is easy. Hopefully JVMs + JNI
doesn't depend upon this char[], 'though it wouldn't
surprise me. If you're concerned about holding onto
substrings, you could treat this as a special case.


REPRODUCIBILITY :
This bug can be reproduced always.

---------- BEGIN SOURCE ----------
Here's a test:

  Thread t = Thread.currentThread();
  while (true) {
     t.setName("foo");
  }

It's easy to make similar tests for "getName()".
---------- END SOURCE ----------

CUSTOMER WORKAROUND :
Suffer the GC overhead. Or not rename the Threads and see
useless thread dumps.
(Review ID: 163857)
======================================================================