question about diamond spec

[Vicente-Arturo Romero-Zaldivar]

Hi Alex,

Thanks,

Vicente

On 06/09/13 18:12, Alex Buckley wrote:
> Hi Vicente,
>
> The JLS specifies only that a compile-time error occurs; it says 
> nothing about diagnostics accompanying the error.
>
> Alex
>
> On 9/6/2013 2:33 AM, Vicente-Arturo Romero-Zaldivar wrote:
>> Hi Alex,
>>
>> We have been working with a javac crash [1]. The underlying reason for
>> the crash is the different diagnostic reports we have in javac for
>> constructor invocations using diamond expressions compared to those with
>> type arguments:
>>
>> List<String> l = new ArrayList<>(); goes through a path
>> different to:
>> List<String> l = new ArrayList<String>();
>>
>> I assume this is to provide more detailed, user friendly, error
>> diagnostics in case a diamond expression can't be applied, but I would
>> like to know if there is a specific specification issue that justify
>> this approach.
>>
>> We would like to merge the diagnostic report of both expressions as much
>> as possible.
>>
>> Thanks,
>>
>> Vicente.
>>
>> [1] 8024207: javac crash in Flow$AssignAnalyzer.visitIdent