On Tue, 9 Feb 2021 00:04:44 GMT, Naoto Sato <na...@openjdk.org> wrote:
>> The fix looks ok to me. Just an interesting note that this test took such a >> long time to run. Given the amount of locales, limit(30) would reduce the >> time to approximately 1/5 of the time, that would still be about 100,000ms >> according to the bug report. That feels pretty long for a single test case >> that, outside of the test environment, takes only 1s (my local test took 5s >> without limit(30)). >> >> Also, there are a bunch of other tests that also run through all of the >> available locales, I wonder if they also take a long time to run. > > As of JDK15, there are more than 800 locales, so reducing the samples down to > 30 makes it less than 4% of the original (~20,000ms on that machine). But I > agree that the test took too long than my environment too. Could be a very > slow machine. > As to other tests calling available locales, the reason this test took a > long time is that the loop multiplies with the number of time zones which is > also huge (600+), so it would be less likely that other locale tests cause > timeouts. I see. I didn't know locales had gone from 160 to 800! All the more reason to set a limit. Rerun with jdk 16, limit(30) does make it a breeze comparing to that without (which made my machine start to spin). 20,000ms is reasonable on a slower machine with a debug build. ------------- PR: https://git.openjdk.java.net/jdk/pull/2465