[concurrency-interest] Re: SynchronizedLong vs. other locks

Dawid Kurzyniec dawidk at mathcs.emory.edu
Fri Jul 8 11:08:17 EDT 2005

Tim Peierls wrote:

> Tim Peierls wrote:
>> Answering my own question: No, it doesn't. The program below produces 
>> this output on my machine (today):
>> t1 = 1120881600000
>> t2 = 1120867200000
>> a difference of 4 minutes. (Hmm, why four minutes? Only 23 leap 
>> seconds have been inserted since the epoch.)
> Ah, four *hours*! That's the time zone difference. Sorry.
> But can one always count on System.currentTimeMillis() in this way?
Looking at GregorianCalendar.computeFieldsImpl(), it is clear that 
GregorianCalendar ignores leap seconds, so it seems safe to replace it 
with that arithmetics. (BTW. the arithmetics work correctly only for 
positive "currentTimeMillis", e.g. after 1970. I don't suppose it 
matters in the discussed application, though).


More information about the Concurrency-interest mailing list