On Mon, 13 Jun 2022 22:21:18 GMT, Mandy Chung <mch...@openjdk.org> wrote:
>> test/lib/jdk/test/lib/util/ForceGC.java line 44:
>>
>>> 42: */
>>> 43: public static boolean wait(BooleanSupplier booleanSupplier) {
>>> 44: return wait(booleanSupplier, 10000L);
>>
>> For the max waiting time, instead of a fixed value I strongly recommend
>> choosing some base waiting time, say 1 second, and scaling it by the jtreg
>> timeout factor.
>
> I also have similar feedback. The timeout factor can be found in the
> `test.timeout.factor` system property, if present. When running tests with
> -Xcomp in a fastdebug build, the timeout factor is currently configured to
> 10. Setting the base waiting time to 1 second seems reasonable.
Good point! I changed to use 1 second as the base waiting time. I was not
very sure if 1 second is big enough for the GC cleanup, but the testing result
on my laptop and VMs, which are super faster, looks good to me. Please help to
check mach5 result.
-------------
PR: https://git.openjdk.org/jdk/pull/8979
