On Tue, 10 May 2022 17:43:24 GMT, Naoto Sato <na...@openjdk.org> wrote:
>> src/java.base/share/classes/java/util/TimeZone.java line 543:
>>
>>> 541: return new ZoneInfo(totalSecs == 0 ? "UTC" : GMT_ID +
>>> tzid, totalSecs);
>>> 542: } else {
>>> 543: return getTimeZone(tzid, true);
>>
>> Before the change in this PR, we used to prefix `GMT` to (non-custom
>> timezone ids) if the timezone id returned by `ZoneId#getId()` started with
>> the `+` or `-` sign, before calling `getTimeZone(modifiedTzid, true)`.
>> With this change, for `ZoneId`s that aren't `ZoneOffset` instance, we now
>> call `getTimeZone(originalTzid, true)`, without first checking/prefixing the
>> id with `GMT`. Is that an intentional change and would that potentially
>> cause `getTimeZone(String, boolean)` to return a different result?
>
> Yes, it is intentional. The `Time-zone IDs` section in the `ZoneId` class
> description is clear that zone id starting with "+/-" is a `ZoneOffset`
> instance. Other ZoneIds should have offsets with prefix or region-based ids.
I was stumbling on this, too, but it is only ZoneOffset isntances that have
those strange names without GMT/UTC prefix. Default ZoneId instances created
from text strings always have a prefix. If you call ZoneId.of() with an offset
only and no prefix it returns a ZoneOffset instance (see the test).
-------------
PR: https://git.openjdk.java.net/jdk/pull/8606
