hg: valhalla/valhalla: 8221545: [lworld] Preliminary support for alternate denotation of values and boxes

[Brian Goetz]

Let’s separate the language and VM aspects.

In the language, since V <: V?, no cast is needed when assigning a V to a V?.  And no CHECKCAST is needed for the corresponding code generation.

In the other direction, a CHECKCAST is definitely needed going from V? to V; the question of whether the _language_ requires an explicit cast is still up in the air.  I think we should lean on explicit casting for now, but I suspect we’ll slide back to implicit narrowing (with possible NPE).  

This also makes a difference for overload selection.  If I have:

    m(Object o) { }
    m(V? v) { }

and a call

    V v = …
    m(v)

Under the old (V.box) interpretation, we’d prefer the m(Object) overload, since it is applicable without a boxing conversion, and m(V?) would not be.  But under the new (subtyping) interpretation, we should treat them both as applicable, and V? is _more specific_ than OIbject, so we pick m(V?).  



> On Apr 16, 2019, at 10:30 AM, Srikanth <srikanth.adayapalam at oracle.com> wrote:
> 
> On a quick check, I see that on both directions, we have a compiler inserted cast.
> As I understand, assignment of V to V? being a widening conversion, the cast should go away while
> the other direction is still being debated.
> 
> This should be straightforward.
> 
> Srikanth
> 
> On 16/04/19 7:50 PM, Brian Goetz wrote:
>> OK, thanks.  Hopefully the new rules will be simpler, as they lean more heavily on subtyping.
>> 
>> 
>>> On Apr 16, 2019, at 10:18 AM, Srikanth <srikanth.adayapalam at oracle.com> wrote:
>>> 
>>> 
>>> This change set just introduces the new notation doing nothing new about inserting casts.
>>> I will go through the mails discussing this and follow up on a different change set.
>>> 
>>> Thanks
>>> Srikanth
>>> On 16/04/19 7:32 PM, Brian Goetz wrote:
>>>>> This was a request for LW2 - no more boxing.
>>>> Right.  I was wondering whether we got there yet, or whether we’d just change the notation to start.
>>>> 
>>>>> We asked for the subtype relationship for both values and for arrays.
>>>> For arrays, the covariance rule: T <: U imples T[] <: U[] — should hold uniformly.  No need for special exceptions for value arrays.  Since
>>>> 
>>>>     V <: Object
>>>> and
>>>>     V? <: Object
>>>> 
>>>> then both V[] and V?[] are <: Object[].
>>>> 
>>>> 
>>>> 
>>>>> For LW2 - we requested explicit checkcasts on attempts to narrow.
>>>>> 
>>>>> Agreed that there are discussions about where this is all going after LW2 before LW10.
>>>>> 
>>>>> thanks,
>>>>> Karen
>>>>> 
>>>>>> On Apr 16, 2019, at 9:24 AM, Brian Goetz <brian.goetz at oracle.com> wrote:
>>>>>> 
>>>>>> Is the conversion from V to V? still considered a boxing conversion, or is it a widening conversion now?  The language we (currently think we) want is
>>>>>> 
>>>>>>  V <: V?
>>>>>> 
>>>>>> In the other direction, we’re still discussing whether the narrowing conversion is implicit or explicit.
>>>>>> 
>>>>>>> On Apr 16, 2019, at 8:09 AM, Srikanth <srikanth.adayapalam at oracle.com> wrote:
>>>>>>> 
>>>>>>> Notes:
>>>>>>> 
>>>>>>>   - This is the initial commit for supporting V? notation as an alternate for V.box
>>>>>>>   - The old .box and .val notation is not withdrawn yet and can still be used, but we should really move over to the new syntax as soon as possible.
>>>>>>>   - Existing javac tests have been replicated to use the V? syntax.
>>>>>>>   - VBC migration related work is not done yet.
>>>>>>>   - One issue that showed up in the parser implementation is the ambiguity at the point of seeing ?
>>>>>>>      in  v instanceof Type? exp: exp2; I have arranged for this to be processed as a conditional rather
>>>>>>>      than as denoting nullability.
>>>>>>>   - On a long value type name the ? suffix does not really stand out ATM, but I guess IDE colorcoding can address this issue.
>>>>>>>   - There is one failing test in the langtools suite (jdk/jshell/TypeNameTest.java) I will raise a follow up
>>>>>>>     ticket to address this.
>>>>>>>   - More testing is required and will be undertaken in the coming weeks. I request/encourage testing by the interested parties too.
>>>>>>> 
>>>>>>> Thanks!
>>>>>>> Srikanth
>>>>>>> 
>>>>>>> On 16/04/19 5:32 PM, srikanth.adayapalam at oracle.com wrote:
>>>>>>>> Changeset: bf52db1ad4ea
>>>>>>>> Author:    sadayapalam
>>>>>>>> Date:      2019-04-16 17:31 +0530
>>>>>>>> URL:       http://hg.openjdk.java.net/valhalla/valhalla/rev/bf52db1ad4ea
>>>>>>>> 
>>>>>>>> 8221545: [lworld] Preliminary support for alternate denotation of values and boxes
>>>>>>>> 
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/comp/Attr.java
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/comp/AttrContext.java
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/comp/Resolve.java
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/parser/JavacParser.java
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/tree/JCTree.java
>>>>>>>> ! src/jdk.compiler/share/classes/com/sun/tools/javac/tree/TreeCopier.java
>>>>>>>> ! test/langtools/tools/javac/valhalla/lworld-values/BogusIncompatibility.java
>>>>>>>> + test/langtools/tools/javac/valhalla/lworld-values/BoxValCastTest2.java
>>>>>>>> + test/langtools/tools/javac/valhalla/lworld-values/GenericsWithQuestion.java
>>>>>>>> + test/langtools/tools/javac/valhalla/lworld-values/GenericsWithQuestion.out
>>>>>>>> ! test/langtools/tools/javac/valhalla/lworld-values/InnerClassAttributeValuenessTest.java
>>>>>>>> + test/langtools/tools/javac/valhalla/lworld-values/IntercastTest2.java
>>>>>>>> + test/langtools/tools/javac/valhalla/lworld-values/LookupOnLoxTest2.java
>>>>>>>> ! test/langtools/tools/javac/valhalla/lworld-values/Point.java
>>>>>>>> 
>