[foreign-abi] [Rev 03] RFR: 8241504: Expose MemoryLayout annotations/attributes in the public API

[Maurizio Cimadamore]

On Tue, 24 Mar 2020 17:38:20 GMT, Jorn Vernee <jvernee at openjdk.org> wrote:

>> Hi,
>> 
>> This PR exposed MemoryLayout attributes in the public API.
>> 
>> A summary of the changes:
>> - Renamed annotation -> attribute everywhere to avoid confusion with Java language annotations
>> - Made sure that annotations were being serialized as part of a MemoryLayout's ConstantDesc, as well as the alignment
>>   which was previously left out. (The latter also meant changing PaddingLayout::hasNaturalAlignment to always return
>>   `true`, so serializing alignment is skipped for padding). Removed special casing for the abiType attribute. This can
>>   now instead be handled through the generic attribute mechanism.
>> 
>> Thanks,
>> Jorn
>
> Jorn Vernee has updated the pull request incrementally with two additional commits since the last revision:
> 
>  - Merge branch 'Attributes' of https://github.com/JornVernee/panama-foreign into Attributes
>  - Remove ML::attribute(String, Class) helper method.

Some more comments

src/jdk.incubator.foreign/share/classes/jdk/internal/foreign/abi/x64/windows/CallArranger.java line 187:

> 186:             if (type.attribute(VARARGS_ATTRIBUTE_NAME)
> 187:                     .filter(String.class::isInstance)
> 188:                     .map(String.class::cast)

Not convinced by the `filter` here? You get an Optional<Constable>, you do a cast to String and then parse as boolean -
why the filtering? If there's an annotation with that name, but different type, I'd prefer to crash than to end up
doing something wrong. Same applies elsewhere.

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

PR: https://git.openjdk.java.net/panama-foreign/pull/64