Hi Peter,
I think this cleans up all the points raised earlier.
The optimization for enqueuing from the reference queue seems ok to me
and should be
more efficient than the previous implementation but I think Mandy or
Alan should look at it also.
Thanks, Roger
On 2/25/2016 4:17 AM, Peter Levart wrote:
Hi Alan,
On 02/25/2016 09:00 AM, Alan Bateman wrote:
On 25/02/2016 07:24, Peter Levart wrote:
:
I kept the public boolean Cleaner.cleanNextPending() method which
now only deals with enqueued Cleanable(s). I think this method might
still be beneficial for public use in situations where cleanup
actions take relatively long time to execute so that the rate of
cleanup falls behind the rate of registration of new cleanup actions.
I think we need also need to look at the option where this is not
public. I have concerns that it is exposing implementation to some
extent and that may become an attractive nuisance in the future. This
shouldn't be an issue for the NIO buffer usage, we can keep the usage
via the shared secrets mechanism. I think this is what Mandy is
suggesting too.
-Alan.
Sure, no problem. Here's a variant that keeps the
Cleaner.cleanNextPending() method private and exposed via
SharedSecrets to nio Bits but is otherwise equivalent to webrev.06:
http://cr.openjdk.java.net/~plevart/jdk9-dev/removeInternalCleaner/webrev.06priv/
Regards, Peter
