On Dec 10, 2013, at 1:25 PM, Jordan Rose <[email protected]> wrote:

> 
> On Dec 10, 2013, at 11:29 , jahanian <[email protected]> wrote:
> 
>> 
>> On Dec 10, 2013, at 9:20 AM, Jordan Rose <[email protected]> wrote:
>> 
>>> Thanks, this is better. Still have some comments though.
>>> 
>>> On Dec 9, 2013, at 14:04 , Fariborz Jahanian <[email protected]> wrote:
>>> 
>>>> + def err_objc_bridged_related_unknown_method : Error<
>>>> + "you can't convert %0 to %1, without using an existing "
>>>> + "%select{class|instance}2 method for this conversion">;
>>> 
>>> I still don't see what this diagnostic adds. If the framework doesn't have 
>>> a method name for one of the conversion directions, that probably means 
>>> there's no one good way to perform this conversion. There may be no way to 
>>> perform the conversion, although it's strange that they would be marked 
>>> related, then.
>>> 
>>> Also, diagnostics generally don't have "you" except in the fix-it bit ("did 
>>> you mean...").
>>> 
>>> How about just "%0 cannot be directly converted to %1”?
>> 
>> I improved the text. I also want to mention that user may have forgotten to 
>> specify a selector name in the attribute.
> 
> I just don't think that's correct. The person who is seeing the error is most 
> likely not the person who wrote the attribute, in which case a missing 
> selector means that that conversion can't be performed.

I wanted to include the info. why the conversion failed. Just got the wording 
wrong. But, warnings can always be changed at a later date.
In r196977.

- Fariborz

> 
> Maybe we should have people explicitly write "0" or "NO" or something for 
> missing pieces?
> 
> Jordan

_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to