[cfe-commits] [Review request] Adding source location for the template keyword in AST template-id expressions.

[Enea Zaffanella]

If there are no objections to the patch recalled here below, we will
commit it.

Enea.


Il 13/01/2012 19:21, Enea Zaffanella ha scritto:
> Ping.
> 
> Il 09/01/2012 18:32, Enea Zaffanella ha scritto:
>> Hello.
>>
>> The current AST representations for expression nodes that possibly
>> include a template-id (namely, DeclRefExpr and MemberExpr, their
>> DependentScope variants and their Unresolved variants) store source
>> location info for:
>>  - the name qualifier (NestedNameSpecifierLoc);
>>  - the name (DeclarationNameInfo);
>>  - the explicit template arguments (ASTTemplateArgumentListInfo).
>>
>> What was missing is the source location of the (often optional,
>> sometimes mandatory) template keyword that may occur between the name
>> qualifier and the name itself.
>>
>> We would appreciate if someone could review the attached (big, but not
>> really complex) patch, which is meant to add this info so as to improve
>> the source-code accuracy of the AST representation.
>>
>> Below we summarize the relevant changes to help in the review process.
>> Comments are welcome.
>>
>> Enea
>>
>>
>>
>> **** Changes in the AST nodes:
>>
>> We extend ASTTemplateArgumentListInfo so as to allow for storing the
>> additional source location of the template keyword (TemplateKWLoc).
>> The new structure is called ASTTemplateKWAndArgsInfo and replaces the
>> old one in all the expression nodes mentioned above.
>> (Note: function template specializations keep using the old
>> ASTTemplateArgumentListInfo structure).
>>
>> In the expression nodes, the structure is allocated if the expression is
>> provided with the template keyword OR the explicit template argument
>> list; previously, it was not allocated if there was no explicit template
>> argument list.
>>
>> Other changes include: helper functions for querying the new location,
>> support for its (de-)serialization, changes to the pretty printer to
>> exploit the knowledge that there was a template keyword.
>>
>>
>> **** Changes in Parse:
>>
>>  - Added a TemplateKWLoc field to TemplateIdAnnotation;
>>  - Modified ParseUnqualifiedId to also return the source location of the
>> optionally parsed template keyword;
>>  - Modified a few other Parse*Id* helpers and the relevant Sema::ActOn*
>> routines to add TemplateKWLoc as an additional input argument;
>>
>>
>> **** Changes in Sema:
>>
>>  - Many (basically trivial) changes to Sema routines so as to propagate
>> the new TemplateKWLoc input argument down to the code creating the
>> relevant AST nodes. The new input argument has been kept on his own on
>> purpose, since it is not really part of the name qualifier or the name
>> itself. Also changed the relevant routines in TreeTransform.h so as to
>> propagate the info on instantiation.
> _______________________________________________
> cfe-commits mailing list
> cfe-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
>