RFR: JDK-8238838: spurious error message for compact constructors with throws clause

[Vicente Romero]

On 2/13/20 6:40 AM, Maurizio Cimadamore wrote:
> Fix looks good. Note that, by using a name instead of a symbol 
> (although I agree it has to be that way, at least in the parser) means 
> that you will get no name disambiguation support from the rich 
> diagnostic formatter. That is - in a case where you have multiple 
> records with same name (but in different scopes), the new message will 
> show just the record name, while I think using symbols will also add 
> package names to disambiguate things. Not that it will be terribly 
> important in this particular case.
>
> Another option would be to create a specific parser diagnostic - but I 
> don't think we need to go that far.

we could do that in the future if we see it necessary

>
> Maurizio

Thanks for the review,
Vicente

>
> On 13/02/2020 04:43, Vicente Romero wrote:
>> ping
>>
>> On 2/11/20 3:35 PM, Vicente Romero wrote:
>>> Hi,
>>>
>>> Please review the fix for [1] at [2]. The fix is oriented at showing 
>>> a friendlier error message in case a throws clause is declared in a 
>>> compact constructor. We already issue an error for this case for 
>>> canonical constructors so the same support could be expected for 
>>> compact ones. I had to modify the arguments for a generic error 
>>> message used for compact and canonical constructors as the error for 
>>> canonical constructors is shown in Attr while the error for compact 
>>> is shown in the parser. This is because the syntax for compact 
>>> constructors doesn't expect anything between the constructor's name 
>>> and its body. As at parser time there are no symbols in the ASTs I 
>>> changed a parameter for the generic error message from `symbol` to 
>>> `name`. This change has provoked that other error logging related to 
>>> constructors had to be changed.
>>>
>>> Thanks,
>>> Vicente
>>>
>>> [1] https://bugs.openjdk.java.net/browse/JDK-8238838
>>> [2] http://cr.openjdk.java.net/~vromero/8238838/webrev.00/
>>