RFR: JDK-8305507 Add support for grace period before AbortVMOnSafepointTimeout triggers [v2]

[David Holmes]

On Fri, 28 Apr 2023 16:23:49 GMT, Wojciech Kudla <duke at openjdk.org> wrote:

>> src/hotspot/share/runtime/safepoint.cpp line 788:
>> 
>>> 786:   // To debug the long safepoint, specify both AbortVMOnSafepointTimeout &
>>> 787:   // ShowMessageBoxOnError.
>>> 788:   if (AbortVMOnSafepointTimeout && Management::ticks_to_ms(os::elapsed_counter()) > (jlong)AbortVMOnSafepointTimeoutDelay) {
>> 
>> Rather than use the Management API here perhaps just:
>> 
>> if (AbortVMOnSafepointTimeout
>>     && (os::elapsedTime() * MILLIUNITS > AbortVMOnSafepointTimeoutDelay)) {
>> 
>> ?
>
> To my understanding `os::elapsedTime() * MILLIUNITS` will reduce the granularity of the grace period down to a second but I fully agree this is nicer and we don't need millis for this functionality. 
> Thank you for the suggestion, @dholmes-ora

`os::elapsedTime()` returns a double so has fractional seconds.

-------------

PR Review Comment: https://git.openjdk.org/jdk/pull/13386#discussion_r1187079468