Hi Ben,

Thanks for the review. I have made most of the changes, but I have a few 
questions.

>> I think it makes sense because c++11 attributes, for example [[loop 
>> vectorize(4)]], would be verified at this point too.
> 
> Why is that?  Shouldn’t they both be dealt with in Sema?  I don’t like the 
> idea that I wouldn’t get these semantic diagnostics with -fsyntax-only.

Are you thinking this should be put in AnalysisBasedWarnings or something? Can 
you be more specific about where in Sema and what calls it?


>> Index: test/Parser/pragma-loop2.cpp
>> ===================================================================
>> --- test/Parser/pragma-loop2.cpp     (revision 0)
>> +++ test/Parser/pragma-loop2.cpp     (working copy)
>> @@ -0,0 +1,50 @@
>> +// RUN: not %clang %s 2>&1 | FileCheck %s
> 
> This test just highlights why this should be done in Sema if at all possible… 
> Also, you probably don’t need the driver here.

What is the driver?


I will provide another patch when responding to Aaron’s review.

Tyler
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to