[threeten-dev] tzdb compiler on boot-jdk

[Stephen Colebourne]

Looks fine to me.

I guess TimeDefinition would be another easy target for "slimming
down", as would replacing factory methods by constructors. But those
can be done later.
Stephen

On 20 January 2013 08:24, Xueming Shen <xueming.shen at oracle.com> wrote:
> hash() & equals() have been missing from the version of ZoneRules.java for
> the tzdb compiler package. So the dup mechanism is broken, the size of
> tzdb.jar that includes n versions of rules is huge.
>
> The fix is to add these two pair back into ZoneRules.java. The webrev
> below includes more updates, mainly removed the dependency for
> the Month and DayOfWeek and more non-needed methods from couple
> classes.
>
> http://cr.openjdk.java.net/~sherman/jdk8_threeten/zruleDup/
>
> Let me know if you don't like this version and prefers the ZoneRules.java
> change:-)
>
> Sherman
>
>
> On 1/18/2013 1:02 PM, Xueming Shen wrote:
>>
>> On 01/18/2013 12:37 PM, Stephen Colebourne wrote:
>>>
>>> Looks like your last webrev doesn't show files as being moved, just
>>> deleted and re-created.
>>
>>
>> I brought them back to generate the "diff" of those three files. The
>> latest is
>> generated via webrev -N, so a real snapshot of the change.
>>
>> leap second can be brought back if the api/impl need them. Maybe 9:-)
>>
>> The test suggests we have the exactly the same rules.
>>
>> -Sherman
>>
>>> Otherwise, I've no problems with pushing if the functionality matches,
>>> although I am sad to lose the leap seconds.
>>> Stephen
>>>
>>> On 18 January 2013 20:39, Xueming Shen<xueming.shen at oracle.com>  wrote:
>>>>
>>>> webrev has been updated, with some minor changes. Fully tested,
>>>> ready to go.
>>>>
>>>> http://cr.openjdk.java.net/~sherman/jdk8_threeten/tzdb_v2
>>>>
>>>> -sherman
>>>>
>>>>
>>>> On 01/18/2013 12:09 PM, Xueming Shen wrote:
>>>>>
>>>>> A simple TzdbDiff tzdb1.jar tzdb2.jar suggests the tzdb.jar generated
>>>>> by the new version matches the previous one. So I'm confident now
>>>>> to go:-)
>>>>>
>>>>> http://cr.openjdk.java.net/~sherman/jdk8_threeten/tzdb_v2/TzdbDiff.java
>>>>>
>>>>> -Sherman
>>>>>
>>>>> On 01/18/2013 10:32 AM, Xueming Shen wrote:
>>>>>>
>>>>>> Yes, they can be further consolidated, if given more time:-)
>>>>>>
>>>>>> On 01/18/2013 10:08 AM, Stephen Colebourne wrote:
>>>>>>>
>>>>>>> This looks like a reasonable approach. Ultimately, these classes
>>>>>>> could
>>>>>>> be slimmed down more, but that isn't vital now.
>>>>>>> Stephen
>>>>>>>
>>>>>>> On 18 January 2013 17:51, Xueming Shen<xueming.shen at oracle.com>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>> Current approach to build tzdb data is a no-go from build and
>>>>>>>> embedded
>>>>>>>> team.
>>>>>>>>
>>>>>>>> Here is the webrev that to have the tzdb compiler works with a set
>>>>>>>> of
>>>>>>>> strip-down
>>>>>>>> JSR310 class to generate the tzdb.jar on the boot_jdk, which is
>>>>>>>> assumed
>>>>>>>> to
>>>>>>>> be
>>>>>>>> a early jdk release that does not have any jdk8 feature. I also
>>>>>>>> remove
>>>>>>>> the
>>>>>>>> leak
>>>>>>>> second from tool and the date.
>>>>>>>>
>>>>>>>> The code probably need further clean-up.  But I will spend most of
>>>>>>>> the
>>>>>>>> time
>>>>>>>> today
>>>>>>>> to make sure the correctness of the output (the tests are all past,
>>>>>>>> but
>>>>>>>> I
>>>>>>>> will verify
>>>>>>>> the output is identical to the output from the "normal" version)
>>>>>>>>
>>>>>>>> http://cr.openjdk.java.net/~sherman/jdk8_threeten/tzdb_v2/
>>>>>>>>
>>>>>>>> -Sherman
>>>>>>
>>>>>>
>>
>