================
Comment at: lib/Parse/Parser.cpp:926-929
@@ +925,6 @@
+ return false;
+ for (unsigned I = 0, E = D.getNumTypeObjects(); I != E; ++I) {
+ unsigned chunkIndex = E - I - 1;
+ const DeclaratorChunk &DeclType = D.getTypeObject(chunkIndex);
+ if (DeclType.Kind == DeclaratorChunk::Function) {
+ const DeclaratorChunk::FunctionTypeInfo &FTI = DeclType.Fun;
----------------
Richard Smith wrote:
> Can you add some more difficult test cases for this (where the order in which
> we visit the objects matter), like:
>
> auto F() -> auto * { return (int*)0; }
> auto (*G())() -> int * { return F; }
Done.
================
Comment at: lib/Parse/Parser.cpp:1014-1015
@@ -994,3 +1013,4 @@
TemplateInfo.Kind == ParsedTemplateInfo::Template &&
- !D.getDeclSpec().isConstexprSpecified()) {
+ !D.getDeclSpec().isConstexprSpecified() &&
+ !isFunctionDeclaratorRequiringReturnTypeDeduction(D)) {
MultiTemplateParamsArg
TemplateParameterLists(*TemplateInfo.TemplateParams);
----------------
Richard Smith wrote:
> Can you make this use something like `Sema::canSkipFunctionBody` instead?
I can have the code in ParseCXXInlineMethods use canSkipFunctionBody - but over
here I am not sure how to do that since all i have at this point to work with
is the Declarator (the Action.HandleDeclarator has not been called - can I call
this early, won't it interfere with the ActOnStartofFunction[Template]Def?
http://llvm-reviews.chandlerc.com/D2053
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
