Bob Hanson wrote:
> Nicolas Vervelle wrote:
>
>> Bob Hanson wrote:
>>
>>
>>> pps. The following is more elegant and also works:
>>>
>>> static Boolean[] testInit = new Boolean[1];
>>> private void initialize() {
>>> if (testInit[0] != null) {
>>> while (!testInit[0].booleanValue()) {
>>> try{
>>> Thread.sleep(10);
>>> } catch(Exception e) {
>>> System.out.println("geodesic constructor error "+e);
>>> }
>>> }
>>> return;
>>> }
>>> testInit[0] = Boolean.FALSE;
>>> ...
>>> System.out.println("geodesic initialization completed");
>>> testInit[0] = Boolean.TRUE;
>>> }
>>>
>>> It takes advantage of the fact that static testInit[0] has THREE
>>> possible values: null, false, and true.
>>> This is what I'm going with for now.
>>>
>>>
>>
>> That won't work in every situation :
>> - thread 1 executes "if (testInit[0] != null)" and finds that it
>> evals to false
>>
>>
> No, you are wrong, I think Thread 1 evaluates it as "null".
> guaranteed. That's a "Boolean" not "boolean".
That's what I was saying: testInit[0] = null so the != evals to false
At the beginning of the program, testInit[0] = null
The first thread that calls initialize() evals (testInit[0] != null) to
false, so it goes on with testInit[0] = Boolean.TRUE
The problem is that the test + assignment is not an atomic operation so
before the assignment is done, an other thread can call the same test
and also get a false, so it goes on with testInit[0] = Boolean.TRUE
Really, check for the Singleton pattern and how it can be correctly
handled for multi-threaded environment.
It's not simple, and it really requires some synchronization at one point
_______________________________________________
Jmol-developers mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/jmol-developers
