[foreign-memaccess] [Rev 01] RFR: JDK-8243064: Add more support for mapped memory segments

[Jorn Vernee]

On Mon, 20 Apr 2020 17:09:29 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

>> This patch adds more targeted support for mapped segments, by providing the same set operations also offered by
>> MappedByteBuffer (force/load/isLoaded).
>> To get there some refactoring of the undelying mapped routines was necessary so that it could be shared across
>> MappedByteBuffer and MappedMemorySegment.
>> One notable consequence of this work is that, in order to keep the typing sharp, I had to turn the
>> MemorySegment::spliterator method from instance to static, so that I could take advantage of generic method type
>> variables. With an instance method it is not possible to write a covariant override that makes sense for clients (note:
>> returning `Spliterator<? extends MemorySegment>` is possible in principle, but useless in practice, as calls to e.g.
>> `tryAdvance` will fail). Overall, I think that is a decent compromise.  Here's a link to the javadoc:
>> http://cr.openjdk.java.net/~mcimadamore/panama/8243064/javadoc/jdk/incubator/foreign/package-summary.html
>> The doc for the new memory mapped segment heavily borrows from MappedByteBuffer - I did only minor tuning here and
>> there to make the doc fit better within the memory segment javadoc.
>> The testing side is thin/non-existing. Unfortunately I could not find any relevant JDK test for mapped buffers that
>> stressed the behavior of the various routines in a meaningful way. For instance, there's a test
>> `MappedByteBuffer/Force.java` which doesn't really test anything, other than calling force() doesn't fail with some
>> error. I've added a call to MappedMemorySegment::force in the buffer segment test, but I realize that this is pretty
>> ad-hoc. If anybody has ideas on how to test this thing in a reliable way, suggestions are welcome :-)
>
> Maurizio Cimadamore has updated the pull request incrementally with one additional commit since the last revision:
> 
>   Add unload() API point to mark mapped segment as 'no longer used'.

Marked as reviewed by jvernee (Committer).

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

PR: https://git.openjdk.java.net/panama-foreign/pull/115