[code-reflection] RFR: Add support for enclosing class types

[Paul Sandoz]

On Thu, 30 May 2024 21:05:15 GMT, Maurizio Cimadamore <mcimadamore at openjdk.org> wrote:

>>> Hmm... i suspect the it's due to canonicalization. We parse the expected model and serialize it out. Depending on the operation its result type is determined by say its operands rather than what is explicitly declared in the text.
>> 
>> That might explain why I cannot reproduce compiling the test and dumping out the IR... but it used to work fine, so I doubt it's the op itself. There must be something flaky in the serialization/deserialization process that's leaving `[]` behind.
>
>> > Hmm... i suspect the it's due to canonicalization. We parse the expected model and serialize it out. Depending on the operation its result type is determined by say its operands rather than what is explicitly declared in the text.
>> 
>> That might explain why I cannot reproduce compiling the test and dumping out the IR... but it used to work fine, so I doubt it's the op itself. There must be something flaky in the serialization/deserialization process that's leaving `[]` behind.
> 
> This kind of give me an idea to add a test that does exactly this: takes a type, serialize it, deserialize it back, and check it's the same thing.

Good idea!

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

PR Review Comment: https://git.openjdk.org/babylon/pull/107#discussion_r1621433107