On Thu, 29 Oct 2020 08:41:04 GMT, Sergey Bylokhov <s...@openjdk.org> wrote:
> The test checks that the timestamp of the mouse event accurately represents > the current time of the system, even if the time of the system is jumped to > the past/future. > > On Unix in xawt we calculate the timstamp using this method: > reset_time_utc = System.currentTimeMillis() - getCurrentServerTime(); > timstamp = reset_time_utc + server_offset; > > (1.) Note that the "server_offset" - timstamp when the native event was > generated, and we try to convert it to the UTC timestamp. Unfortunatly we > callculate the "reset_time_utc" only once at the start of the application and > then rarely update it. So if the time in the system is changed we still use > the old one. > > The exactly the same bug described at (1.) was fixed on windows by the > https://bugs.openjdk.java.net/browse/JDK-6461933 and for that bug the test in > question was created. That bug was fixed by the "recalculation" system time > more often. But it does not solve the general time issue and the code was > reworked again by the https://bugs.openjdk.java.net/browse/JDK-8046495 > > I would like to fix the current bug in the same was as on windows, see link > below: > https://bugs.openjdk.java.net/browse/JDK-8046495?focusedCommentId=13517205&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-13517205 > > After the fix we will use the same > System.currentTimeMillis()/JVM_CurrentTimeMillis on all platforms. Any volunteers for a review? ------------- PR: https://git.openjdk.java.net/jdk/pull/925
