8027848: The ZoneInfoFile doesn't honor future GMT offset changes
xueming.shen at oracle.com
Wed Nov 6 17:18:57 UTC 2013
On 11/05/2013 03:21 PM, Aleksej Efimov wrote:
> Thank you for a quick review. I totally agree with you on all items.
> Actually, I missed fact that the transitions are sorted. And yes - the change can be done on line #431.
> The new tested fix can be found here: http://cr.openjdk.java.net/~aefimov/8027848/webrev.01/ <http://cr.openjdk.java.net/%7Eaefimov/8027848/webrev.01/>
> On 11/05/2013 10:58 PM, Xueming Shen wrote:
>> On 11/05/2013 10:50 AM, Xueming Shen wrote:
>>> For better performance
>>> (1) the currT should be "static final" so we dont have to access the
>>> System.curentTimeMillis() for each TimeZone/ZoneInfo instance.
>>> (2) instead of iterating through the standardTransitions(), shouldn't we just
>>> check the last one? given it's a sorted list.
>> and maybe this willGMTOffsetChange setting can be done just at line#431.
>>> btw, in theory, the change now uses the "current vm starttime" instead of
>>> the "tzdb generated" time. But it should be fine, given ZoneInfo.getRawOffset()
>>> will just do a search for the current rawoffset. I may consider to add the
>>> "generated time" into the tzdb.dat in the future, if desired.
>>> On 11/05/2013 09:26 AM, Aleksej Efimov wrote:
>>>> Can I have a review for a 8027848  bug fix . There is unimplemented functionality related to the future GMT offset changes.
>>>> The ZoneInfoFile class doesn't analyses if there is such offset changes and as a result incorrectly creates the ZoneInfo instance.
>>>> It was discovered during the TestZoneInfo310 test execution as part of tzdata2013h update .
>>>>  The bug: https://bugs.openjdk.java.net/browse/JDK-8027848
>>>>  Proposed fix: http://cr.openjdk.java.net/~aefimov/8027848/webrev.00
>>>>  tzdata2013h review thread: http://mail.openjdk.java.net/pipermail/core-libs-dev/2013-November/022898.html
More information about the core-libs-dev