================
Comment at: lib/Sema/SemaDeclAttr.cpp:655-656
@@ +654,4 @@
+ if (!isa<FunctionDecl>(D) && !isa<FunctionTemplateDecl>(D)) {
+ S.Diag(Attr.getLoc(), diag::warn_attribute_wrong_decl_type)
+ << Attr.getName() << ExpectedFunctionOrMethod;
+ return;
----------------
Kostya Serebryany wrote:
> Dmitri Gribenko wrote:
> > Can we promote this to an error? There's no existing code that (mis)uses
> > this attribute, so there should be no compatibility concerns.
> >
> > Please also add tests for this warning or error, and for
> > checkAttributeNumArgs() condition above.
> I copy-pasted the code from the pre-existing code for
> no_thread_safety_analysis
> and I did not fund such tests for that one.
> I guess we need tests for all 3 (+ no_address_safety_analysis).
> Where is the best place for them? Do you know similar existing tests?
>
> Yes, I guess we can make it an error. Will do the change tomorrow.
There are tests for no_address_safety_analysis in
test/SemaCXX/warn-thread-safety-parsing.cpp (lines 93-121).
For this attribute, tests should probably go into
test/SemaCXX/attr-no-uninitialized-checks.cpp (because we don't do static
analysis, it is attr- instead of warn-).
http://llvm-reviews.chandlerc.com/D390
_______________________________________________
cfe-commits mailing list
cfe-commits@cs.uiuc.edu
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
