[foreign-abi] RFR: 8253823: Investigate ways to make handoff-like operation more explicit [v2]

[Maurizio Cimadamore]

On Wed, 30 Sep 2020 09:58:34 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

>> src/jdk.incubator.foreign/share/classes/jdk/incubator/foreign/MemorySegment.java line 909:
>> 
>>> 907:      * @see MemorySegment#handoff(HandoffTransform)
>>> 908:      */
>>> 909:     interface HandoffTransform {
>> 
>> There might be a use-case for starting out with the existing confinement properties of a segment, say if just
>> registering a cleaner or adding an attachment e.g. `HandoffTransform.ofSegment`, which is not a good name since it
>> would imply the acceptance of a segment parameter, whereas it would be applied to the segment that accepts the handoff
>> transform. That way it's not necessary to query a segment before constructing the correct transform.  I suspect there
>> will only be one implementation of `HandoffTransform`, since it's really just a dumb holder of accumulated state, so it
>> could be made into a `final class` (value-based) with a non-public constructor.
>
> `HandoffTransform.identity()` ?

Yes - this is a dumb holder for accumulated state - my question in the review header was as to whether we're fine with
it being dumb and mutable and thread unsafe, since that is reflected in the javadoc

-------------

PR: https://git.openjdk.java.net/panama-foreign/pull/361