CC'ing in Chris, who had expressed concerns in the original PR.

On Wed, May 27, 2015 at 12:51 PM, Alexey Frolov
<[email protected]> wrote:
>> Aaron Ballman wrote:
>
>> "I am unfamiliar with __attribute__((mode)); is there documentation that 
>> supports this assertion, or popular code in the wild that is relying on it? 
>> GCC's docs are very spartan on the >topic, and there was an (unanswered) 
>> question in the PR regarding why we would want to support this."
>
> __attribute__((mode)) is a GCC extension. It is already supported in Clang, 
> but only partially as the current implementation does not cover cases with 
> base vector types.
> I tend to claim this issue (described in PR17453) a missing 
> feature/incomplete implementation as GCC compiler does support base vector 
> types.
> The community already accepted similar patches for this attribute (issues 
> described in PR16752, PR8703 and PR3691, for example). I hope to think this 
> patch is desirable too.

I am familiar with that. What I am unfamiliar with is documentation
about GCC's implementation so we can be sure we're matching their
semantics, or popular source code relying on that functionality. I'm
not opposed to the patch, just want to make sure it's worth the burden
of supporting it.

~Aaron

>> Aaron Ballman wrote: "The diagnostic you added does not accept any 
>> arguments."
>
>>
>
>>   +    if (ComplexMode) {
>
>>   +      S.Diag(Attr.getLoc(), diag::err_complex_mode_vector_type) << Name;
>
>>
>
>> Aaron Ballman wrote: "This diagnostic also does not accept any arguments."
>
>>
>
>>   +  if (NewTy.isNull()) {
>
>>   +    S.Diag(Attr.getLoc(), diag::err_mode_wrong_type) << Name;
>
>
> I removed arguments for the two mentioned diagnostics and updated the patch.
>
>
> REPOSITORY
>   rL LLVM
>
> http://reviews.llvm.org/D10058
>
> Files:
>   include/clang/Basic/DiagnosticSemaKinds.td
>   lib/Sema/SemaDeclAttr.cpp
>   test/CodeGen/attr-mode-vector-types.c
>   test/Sema/attr-mode-vector-types.c
>
> EMAIL PREFERENCES
>   http://reviews.llvm.org/settings/panel/emailpreferences/

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

Reply via email to