RFR: 8323023: GenShen: Region logging test fails intermittently

[Kelvin Nilsen]

On Fri, 5 Jan 2024 10:41:44 GMT, Aleksey Shipilev <shade at openjdk.org> wrote:

>> This test used to exercise a specific Shenandoah feature, but we've long since rewritten that feature to be based on unified logging. The test doesn't need to retest unified logging features.
>
> test/hotspot/jtreg/gc/shenandoah/TestShenandoahRegionLogging.java line 47:
> 
>> 45:         for (long c = 0; c < count; c++) {
>> 46:             sink = new Object();
>> 47:             Thread.sleep(1);
> 
> Pre-existing, but: The problem with doing this sleep is that we are doing about 64K of 1ms sleeps in this tests. Even if we manage to do 1ms sleep, that is already 65 seconds spent in this test. But we might as well spend single-digit ms per sleep. For 10ms sleeps, that is 10 minutes for a test, which would be playing it close to timeout.
> 
> So I wonder if this test should be time-bound.

Would it be better to reduce the frequency of the sleep, to one in 10, for example?

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

PR Review Comment: https://git.openjdk.org/shenandoah/pull/380#discussion_r1442947910