Re: JEP 370 - text example leads to exception

[Maurizio Cimadamore]

Thanks Chris, both for catching the mistake and for testing out the 
feature - as you can imagine we'd like to collect as much feedback as 
possible to make sure the API gets finalized right.

Cheers
Maurizio
On 11/02/2020 02:12, Chris T wrote:
Paul, thank you very much!

   In the meantime I did more reading about VarHandles and understood better
the "philosophy" of coordinates and it made sense (eventually I was able to
fix my own issues).
   Also, thank you for sharing Maurizio's talk - I will refer it in my
examples. Speaking of sharing - these days I will finish my examples on JEP
370 and will share them with you as well (there will be a github + some
youtube introduction to the feature).
   I think that the capability to allocate more than 2GB of memory is going
to be a hit!

Cheers!
   Chris T

On Mon, Feb 10, 2020 at 12:58 PM Paul Sandoz <paul.san...@oracle.com> wrote:

Thanks for pointing out the inconsistencies.

I modified the JEP with updated code snippets that compile against the
latest API in JDK 14 [*].

The handle created “withStride” requires an additional coordinate the is
an offset from the base address.

You may find Maurizio’s recent talk at Fosdem 2020 helpful and informative:

   https://fosdem.org/2020/schedule/event/bytebuffers/

Hth,
Paul.

[*] I wish there was a way to automate the compile and test of such
snippets without duplication.


On Feb 7, 2020, at 7:23 PM, Chris T <tech.mesh...@gmail.com> wrote:

I tried to build an example on top of this code snippet (from the JEP
text):

VarHandle intHandle = MemoryHandles.varHandle(int.class);
VarHandle intElemHandle = MemoryHandles.withStride(intHandle, 4);

try (MemorySegment segment = MemorySegment.allocateNative(100)) {
   MemoryAddress base = segment.baseAddress();
   for (int i = 0 ; i < 25 ; i++) {
        intElemHandle.set(base, (long)i);
   }
}

The first issue was that the API for the first line need to get the
ByteOrder parameter (that's not a big deal). I ended up with this:

    VarHandle intHandle = MemoryHandles.varHandle(int.class, order);
    VarHandle intElemHandle = MemoryHandles.withStride(intHandle, 4);

    try (MemorySegment segment = MemorySegment.allocateNative(100)) {
      MemoryAddress base = segment.baseAddress();
      for (int i = 0; i < 25; i++) {
        // this is the line where the app crashes:
        intElemHandle.set(base, (long) i);
      }
    }

The issue that I have is that the code crashes with:
java.lang.invoke.WrongMethodTypeException: cannot convert
MethodHandle(VarHandle,MemoryAddressProxy,long,int)void to
(VarHandle,MemoryAddress,long)void
at

java.base/java.lang.invoke.MethodHandle.asTypeUncached(MethodHandle.java:880)
at java.base/java.lang.invoke.MethodHandle.asType(MethodHandle.java:865)
at

java.base/java.lang.invoke.VarHandleGuards.guard_LJ_V(VarHandleGuards.java:184)
at

com.github.kbnt.java14.fma.ForeignMemoryAccessExamples.exampleXXStrides(ForeignMemoryAccessExamples.java:65)
at

com.github.kbnt.java14.fma.ForeignMemoryAccessExamples.main(ForeignMemoryAccessExamples.java:25)

Any idea why this happens (and more importantly how the code can be
changed)?

Thanks!
  Chris T