RFR: 8210142: java.util.Calendar.clone() doesn't respect sharedZone flag
roger.riggs at oracle.com
Tue Sep 4 20:41:40 UTC 2018
That access via reflection is going to go away sometime; so I'm not too
maintaining compatibility of the internal implementation.
I think I'd rather see the memory savings, however small.
Let see if anyone else has a recommendation.
On 9/4/18 4:12 PM, Naoto Sato wrote:
> Hi Roger,
> Yes, I considered that too, but did not change the behavior and just
> to maintain the field consistent. I agree that it would not be
> observable via the public Calendar API, but some apps (like how the
> submitter found it) may be doing something using reflection.
> On 9/4/18 12:31 PM, Roger Riggs wrote:
>> Hi Naoto,
>> The spec for clone doesn't say whether the clone should share or not
>> share the TimeZone.
>> Did you consider that if sharedZone was true , *not* to clone the
>> It would still get cloned when requested from getTimeZone().
>> This does seem somewhat safer not to change the cloning behavior but
>> I don't think the behavior would be observable.
>> The current code and test is fine, except for reducing the potential
>> for sharing the TimeZone.
>> Thanks, Roger
>> On 9/4/2018 2:14 PM, Naoto Sato wrote:
>>> Please review the fix to the following issue:
>>> The proposed fix is located at:
>>> The fix is a simple one line change, which is to make the sharedZone
>>> field consistent with the cloned TimeZone instance in Calendar.clone().
More information about the core-libs-dev