We recently updated tzdata to 2009q in the ICU4J trunk. This change triggers one test case failure - TimeZoneRegression/TestT5280. The test case verifying the behavior of setRawOffset, which no longer make much sense for time zone with historic/irregular offset transitions. For now, setRawOffset artificially creates SimpleTimeZone object close enough to represent current date, then change the raw offset. To create the SimpleTimeZone instance, BasicTimeZone#getSimpleTimeZoneRulesNear is used internally. But, this API returns a pair of TimeZoneRules with different raw offset for Asia/Novokuznetsk in 2009q. These detected rules are used for constructing SimpleTimeZone which results DST savings to 0. (Nearest transition for Asia/Novokuznetsk change STD->DST and raw offset at the same time, which confuse getSimpleTimeZoneRulesNear).
Milestone 4.3.3 deleted