Javac compiler error discovered via project Kulla REPL?

[Alex Buckley]

OK, so in the method invocation expression (whether standalone or on the 
RHS of an assignment), javac should classify the identifier "Undefined" 
as a PackageName -- see JLS 6.5.2. Since no package with that name 
exists, an error is due - see JLS 6.5.3.1. It would be helpful for the 
"cannot find symbol" message to indicate that neither a variable nor a 
type nor a package called "Undefined" is in scope.

I expect that when javac's second error message improves, so will the 
REPL's third error message.

Alex

On 4/16/2015 12:57 PM, Robert Field wrote:
> Actually, none of the below.
>
> They both give the same errors: both a variable ClassName and a class
> ClassName error.  Which makes it even more potentially confusing.
>
>  From Jan's bug report (showing the behavior of javac (no REPL)) (I've
> added bolding) --
>
>     Consider this code:
>     ---
>     public class Test {
>           private void test() {
>               Undefined x = Undefined.create();
>           }
>     }
>     ---
>
>     When compiled, this code produces the following errors:
>     ---
>     Test.java:3: error: cannot find symbol
>               Undefined x = Undefined.create();
>               ^
>        symbol: *class Undefined*
>        location: class Test
>     Test.java:3: error: cannot find symbol
>               Undefined x = Undefined.create();
>                             ^
>        symbol: *variable Undefined*
>        location: class Test
>     2 errors
>     ---
>
> The REPL does --
>
>     -> Undefined x = Undefined.create()
>     |  Error:
>     |  cannot find symbol
>     |    symbol:   class Undefined
>     |  Undefined x = Undefined.create();
>     |  ^-------^
>     |  Error:
>     |  cannot find symbol
>     |    symbol:   class Undefined
>     |  Undefined x = Undefined.create();
>     |  ^-------^
>     |  Error:
>     |  cannot find symbol
>     |    symbol:   variable Undefined
>     |  Undefined x = Undefined.create();
>     |                ^-------^
>
>
>     ->
>
> The only differences are how the position is marked and the REPL gives
> the first one twice (because it has been taken apart and put back
> together with three occurrences of the name).
>
> Given just "Undefined.create()"  both javac and REPL give just a
> "variable Undefined" error
>
> -Robert
>
>
> On 04/16/15 11:40, Alex Buckley wrote:
>> The situation seems to be that the REPL gives no error for the
>> identifier "ClassName" on the LHS of the assignment. Therefore, it
>> must have been possible to classify the identifier as a simple name,
>> in particular as a TypeName. Then, the use of the TypeName on the RHS
>> of the assignment cannot be an error.
>>
>> I can imagine an alternative situation: the identifier "ClassName" is
>> totally made up, and the REPL just happens to be reporting only one of
>> the two "cannot find symbol" errors indicated by JDK-8077970.
>>
>> Which situation is it?
>>
>> Alex
>>
>> On 4/16/2015 8:07 AM, Jan Lahoda wrote:
>>> Hello Martijn,
>>>
>>> I've filled:
>>> https://bugs.openjdk.java.net/browse/JDK-8077970
>>>
>>> Thanks,
>>>      Jan
>>>
>>> On 16.4.2015 13:35, Martijn Verburg wrote:
>>>> Hi all,
>>>>
>>>>  From the REPL hackday we had recently:
>>>>
>>>> "Following error message might be confusing, as the symbol might not
>>>> be a
>>>> variable:
>>>> -> ClassName x = ClassName.create()
>>>> |  cannot find symbol
>>>> | symbol:   variable ClassName
>>>> |  ClassName x = ClassName.create();
>>>> |                ^-------^
>>>> "
>>>>
>>>> Kulla team suggested this is a Java compiler (javac) error message
>>>> and it
>>>> should be submitted as a bug report against javac.
>>>>
>>>> Cheers,
>>>> Martijn
>>>>
>