[cfe-commits] r67823 - /cfe/trunk/lib/Sema/SemaType.cpp

[Douglas Gregor]

On Mar 26, 2009, at 11:15 PM, Chris Lattner wrote:

>
> On Mar 26, 2009, at 11:01 PM, Douglas Gregor wrote:
>
>>
>> On Mar 26, 2009, at 10:50 PM, Chris Lattner wrote:
>>
>>>
>>> On Mar 26, 2009, at 10:10 PM, Douglas Gregor wrote:
>>>
>>>> Author: dgregor
>>>> Date: Fri Mar 27 00:10:56 2009
>>>> New Revision: 67823
>>>>
>>>> URL: http://llvm.org/viewvc/llvm-project?rev=67823&view=rev
>>>> Log:
>>>> Remove the code insertion hint for implicit int. Too often, we're  
>>>> wrong about this hint, so it loses its usefulness. Maybe some day  
>>>> we can make the hint smart enough to be useful.
>>>
>>> I don't get it, is the problem that the insertion point was wrong,  
>>> or that this was the first of a series of errors and the parser  
>>> was just confused?
>>
>> It's the first of a series of errors. I think the root of the  
>> problem is that we emit the "implicit int" warning well before  
>> we've actually done the rest of the type checking for this  
>> declarator.
>
> Ok, well that's kind of the design of the fixit advice:  
> theoretically and IDE would only show fixit advice if all generated  
> errors had fix it's.  If any errors were produced without fixit  
> advice, then we'd assume that error recovery got confused and that  
> none of the fixit advice is trustworthy.

I'm about 85% convinced about that. For me, it depends on just how  
accurate our fix-it advice turns out to be, and I'm still going to  
argue that if we can't be reasonably accurate, we shouldn't suggest  
anything.

>
> There is a problem here with the command line interface, in that we  
> stream out diags as we see them.  We can't "unprint" fixit advice  
> retroactively.

Right.

>
>> That gives me an idea: why not just make a note that we're  
>> potentially dealing with an implicit int, and then only emit this  
>> warning/error after we've successfully processed the declarator?  
>> Then, we omit the diagnostic for semantic disasters, and when we do  
>> provide the diagnostic, the code insertion hint is likely to be  
>> right.
>
> That would be nice refinement in any case, I guess, but the meta  
> issue stands: fixit advice is only trustworthy when the front-end  
> isn't hopelessly confused.


I agree. Perhaps it's just a question of whether we suppress the fix- 
it advice from the front end itself, or whether we rely on the  
diagnostic engine to do it. In an IDE, we're assuming that the  
suppression will happen after the front-end has completed its execution.

   - Doug