[foreign-memaccess+abi] RFR: 8265760: Memory segments/addresses should have an isNative predicate

[Maurizio Cimadamore]

On Fri, 23 Apr 2021 15:36:47 GMT, Paul Sandoz <psandoz at openjdk.org> wrote:

>> This patch adds an `isNative` predicate to both `MemorySegment` and `MemoryAddress`. Note that, for segments, `isMapped` implies `isNative`.
>> 
>> It also moves all the methods from `MappedMemorySegments` back into `MemorySegment` (these methods will still throw if invoked on a segment that is not a mapped segment).
>> 
>> Some javadoc tweaking occurred here - MemorySegment now has a new section fully dedicated to mapped segments (note that now mapped segments are presented as a special class of native segments).
>> 
>> Also, in MemoryAddress, the concept of "native memory address" is introduced to talk about off-heap addresses, which simplifies some of the javadoc.
>> 
>> This way, the resulting API is tighter, and using mapped segments is slightly more direct.
>
> src/jdk.incubator.foreign/share/classes/jdk/incubator/foreign/MemorySegment.java line 94:
> 
>> 92:  * <p>
>> 93:  * Clients requiring sophisticated, low-level control over mapped memory segments, should consider writing
>> 94:  * custom mapped memory segment factories; using JNI, e.g. on Linux, it is possible to call {@code mmap}
> 
> Can we replace the reference to JNI with the linker API?

Sure - that's actually what I wanted to do in the first place - dunno how I ended up with JNI :-)

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

PR: https://git.openjdk.java.net/panama-foreign/pull/514