JVM64bit running on Linux 64bit: when system time changes the JVM may hang (bug_id=6900441)

Andrew Haley aph at redhat.com
Tue Sep 3 00:45:50 PDT 2013


On 09/02/2013 11:04 PM, David Holmes wrote:
> I strongly believe in two distinct notions of time: relative and 
> absolute. Any API involving relative time should never be affected by 
> changes to the absolute value of a clock; and anything involving 
> absolute times obviously should.

I think that seems sensible.  It may not be hard to come up with
a strawman patch for this.

However, while this will fix relative-timed queues, there will be
lurking bugs elsewhere.  These don't even need to be in Java: perhaps
there is code elsewhere that checks filesystem timestamps.  That's not
even unusual.  There's cron, there's make, and any number of other
things.  Maybe such things will be mission-critical.

So we can't say to people that it's OK to move the system clock
backwards.  The best we can say is that it won't cause relative-timed
queues to malfunction.  And the best advice remains "Don't do that."

Andrew.


More information about the hotspot-runtime-dev mailing list