FULL PRODUCT VERSION :
java version "1.5.0_10"
Java(TM) 2 Runtime Environment, Standard Edition (build 1.5.0_10-b03)
Java HotSpot(TM) Client VM (build 1.5.0_10-b03, mixed mode, sharing)

ADDITIONAL OS VERSION INFORMATION :
Window XP

A DESCRIPTION OF THE PROBLEM :
SimpleDateFormat use current timezone offset when parsing a date with Timezone and not the offset of the timezone at the parsed date.

See : method SDF.subParseZoneString when tz != null => use of tz.getRawOffset().

A consequence of this is that date before 1940 with tz Europe/Paris "loose" one hour each time it's parsed :

INFO: Date_object_1 init : Tue Feb 11 00:00:00 CET 1936
INFO: Date_str_1 formated from Date_object_1 : 1936-02-11 00:00:00.0 CET
INFO: Date_object_2 parsed from Date_str_1 : Mon Feb 10 23:00:00 CET 1936
INFO: Date_str_2 formated from Date_object_2 : 1936-02-10 23:00:00.0 CET
INFO: Date_object_3 parsed from Date_str_2 : Mon Feb 10 22:00:00 CET 1936



STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
Set host TZ to Europe/Paris and execute testcase provided.


REPRODUCIBILITY :
This bug can be reproduced always.

---------- BEGIN SOURCE ----------
public class DateFormatTest extends TestCase {

Logger logger = Logger.getLogger("");

public void testParseDateBefore1940Cet() throws ParseException {
String dateStr = "1936-02-11";

DateFormat df1 = new SimpleDateFormat("yyyy-MM-dd");
df1.setTimeZone(TimeZone.getTimeZone("Europe/Paris"));
Date d1 = df1.parse(dateStr);
logger.info("Date_object_1 init : " + d1.toString());

DateFormat df2 = new SimpleDateFormat("yyyy-MM-dd HH:mm:ss.S zzz");
df2.setTimeZone(TimeZone.getTimeZone("Europe/Paris"));

String str = df2.format(d1);
logger.info("Date_str_1 formated from Sate_object_1 : " + str);

d1 = df2.parse(str);
logger.info("Date_object_2 parsed from Sate_str_1 : " + d1);

str = df2.format(d1);
logger.info("Date_str_2 formated from Sate_object_2 : " + str);

d1 = df2.parse(str);
logger.info("Date_object_3 parsed from Date_str_2 : " + d1);

}

}
---------- END SOURCE ----------