r186128 - Provide a fixit hint for changing '->' to '.' if there is no operator->

[Richard Smith]

On Fri, Jul 12, 2013 at 3:16 PM, Kaelyn Uhrain <rikka at google.com> wrote:
> On Thu, Jul 11, 2013 at 10:16 PM, Nico Weber <thakis at chromium.org> wrote:
>>
>> Nice, thanks!
>>
>> On Thu, Jul 11, 2013 at 3:38 PM, Kaelyn Uhrain <rikka at google.com> wrote:
>>>
>>> Author: rikka
>>> Date: Thu Jul 11 17:38:30 2013
>>> New Revision: 186128
>>>
>>> URL: http://llvm.org/viewvc/llvm-project?rev=186128&view=rev
>>> Log:
>>> Provide a fixit hint for changing '->' to '.' if there is no operator->
>>> defined for a class.
>>>
>>> Modified:
>>>     cfe/trunk/lib/Sema/SemaOverload.cpp
>>>     cfe/trunk/test/FixIt/fixit.cpp
>>>
>>> Modified: cfe/trunk/lib/Sema/SemaOverload.cpp
>>> URL:
>>>
http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Sema/SemaOverload.cpp?rev=186128&r1=186127&r2=186128&view=diff
>>>
>>>
==============================================================================
>>> --- cfe/trunk/lib/Sema/SemaOverload.cpp (original)
>>> +++ cfe/trunk/lib/Sema/SemaOverload.cpp Thu Jul 11 17:38:30 2013
>>> @@ -11353,10 +11353,17 @@ Sema::BuildOverloadedArrowExpr(Scope *S,
>>>      break;
>>>
>>>    case OR_No_Viable_Function:
>>> -    if (CandidateSet.empty())
>>> -      Diag(OpLoc, diag::err_typecheck_member_reference_arrow)
>>> -        << Base->getType() << Base->getSourceRange();
>>> -    else
>>> +    if (CandidateSet.empty()) {
>>> +      QualType BaseType = Base->getType();
>>> +      if (BaseType->isRecordType() && !BaseType->isPointerType()) {
>>> +        Diag(OpLoc, diag::err_typecheck_member_reference_suggestion)
>>> +          << BaseType << 1 << Base->getSourceRange()
>>> +          << FixItHint::CreateReplacement(OpLoc, ".");
>>
>>
>> Since this is a fixit on an error, don't you need to fix up the AST?
>> (http://clang.llvm.org/docs/InternalsManual.html#fix-it-hints)
>
>
> This is a tricky case because of how operator-> is recursively applied to
> it's own result... from what I could tell, recovery would require not just
> breaking out of the loop for operator-> in ActOnStartCXXMemberReference
but
> also telling the parser to start over with the '.' in place of the '->',
> which I have no idea how to do from several function-call-layers within
> Sema. And I couldn't add the fixit to a note as the internals manual
> suggests since there is no note associated with the diagnostic.

We usually handle these cases by adding a note to carry the fixit:

error: member reference type 'PR15045::Cl0' is not a pointer
   return c->a;
          ^
note: did you mean to use '.' instead?
   return c->a;
           ~~
           .
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130712/6e653382/attachment.html>