Hi Daniel, it might be a better idea to scale sleep time using jdk.test.lib.Utils::adjustTimeout, so it will remain the same on configuration w/ timeout-factor=1, and scaled proportionally to timeout-factor on slow configurations. the rest looks good.
Thanks, -- Igor > On Jan 24, 2019, at 4:20 AM, Daniel Fuchs <[email protected]> wrote: > > Thanks Lance. > > Meanwhile I noticed that the first loop calling gc() used > a sleep(100) - which is probably not enough in those > greedy configurations - so I extended that to 1000. > (webrev updated in place). > http://cr.openjdk.java.net/~dfuchs/webrev_8217353/webrev.00/ > > Hopefully that will reduce the frequency of failures. > Re testing in progress... > > best regards, > > -- daniel > > On 24/01/2019 12:14, Lance Andersen wrote: >> Morning Daniel, >> The change seems like a good approach >> Best >> Lance >>> On Jan 24, 2019, at 6:42 AM, Daniel Fuchs <[email protected] >>> <mailto:[email protected]>> wrote: >>> >>> Hi, >>> >>> Please find below a fix for: >>> >>> 8217353: >>> java/util/logging/LogManager/Configuration/updateConfiguration/HandlersOnComplexResetUpdate.java >>> >>> fails with Unexpected reference: java.lang.ref.WeakReference >>> https://bugs.openjdk.java.net/browse/JDK-8217353 >>> >>> webrev: >>> http://cr.openjdk.java.net/~dfuchs/webrev_8217353/webrev.00/ >>> >>> This test has been seen failing from time to time with an >>> exception originating from the finally { } clause of the >>> main method. The suspicion is that this exception in fact >>> hides another exception previously thrown in the try { } >>> body. >>> >>> This is only a fix in so far that it will ensure that the >>> real exception - if any - is added to the suppressed list >>> of the secondary exception so that we eventually see it in >>> the log. >>> >>> As to why the test failed in the first place - I am not quite >>> sure, but it might be due to -Xcomp or some other combination >>> of options preventing the WeakReference to be cleared within >>> the time (and number of gc() calls) expected by the test >>> (the test has never been seen failing without these combinations). >>> >>> best regards, >>> >>> -- daniel >> <http://oracle.com/us/design/oracle-email-sig-198324.gif> >> <http://oracle.com/us/design/oracle-email-sig-198324.gif><http://oracle.com/us/design/oracle-email-sig-198324.gif> >> <http://oracle.com/us/design/oracle-email-sig-198324.gif>Lance Andersen| >> Principal Member of Technical Staff | +1.781.442.2037 >> Oracle Java Engineering >> 1 Network Drive >> Burlington, MA 01803 >> [email protected] <mailto:[email protected]> >
