Re: RFR (S) 8205725: Update the JVMTI Spec for Heap Sampling

[serguei.spit...@oracle.com]

Hi all, We need at least one more review before pushing it. Thanks, Serguei On 7/16/18 16:07, JC Beyler wrote: Hi all, The CSR has recently been approved, could someone else review the spec update webrev: http://cr.openjdk.java.net/~jcbeyler/8205725/webrev.03/ The associated bug is here: https://bugs.openjdk.java.net/browse/JDK-8205725 The associated CSR is here: https://bugs.openjdk.java.net/browse/JDK-8206940 Thanks all! Jc On Thu, Jul 12, 2018 at 2:27 PM serguei.spit...@oracle.com <serguei.spit...@oracle.com> wrote: Hi Jc, It looks good to me (including the CSR that I'had already reviewed). Thank you for preparing a fix for this issue so quickly! Thanks, Serguei On 7/12/18 13:45, JC Beyler wrote: Hi all, Could I get a review of an update to the JVMTI Spec for Heap Sampling: http://cr.openjdk.java.net/~jcbeyler/8205725/webrev.03/ The assoicated bug is here: https://bugs.openjdk.java.net/browse/JDK-8205725 The associated CSR is here: https://bugs.openjdk.java.net/browse/JDK-8206940 The basic reasoning of this webrev/bug/CSR is: - rate is not the right word and should be renamed to interval, this is what provokes the change in the code/tests/API naming. - the spec does not mention that the new sampling interval will take time to be taken into account (you have to wait for a TLAB to be refilled); this adds that precision so that the user is not surprised - the spec explicitly says that the sampling is done via a geometric variable which averages to the sampling interval; it was asked to relax this and the spec should just say that the sampling is pseudo-random and the interval will average out to what the user requested. Thanks for all your help, Jc -- Thanks, Jc