[PATCH] Create calling convention AttributedType sugar nodes

[John McCall]

On Jul 19, 2013, at 10:17 AM, Reid Kleckner <rnk at google.com> wrote:
> On Fri, Jul 19, 2013 at 1:04 PM, John McCall <rjmccall at apple.com> wrote:
> On Jul 12, 2013, at 5:39 AM, Reid Kleckner <rnk at google.com> wrote:
> > Canonical types are unchanged.  The type printer had to be changed to
> > avoid printing any non-default implicit calling convention as well as
> > the calling convention attribute.
> 
> +bool AttributedType::isCallingConv() const {
> +  switch (getAttrKind()) {
> +  default:  return false;
> +  case attr_cdecl:
> +  case attr_fastcall:
> +  case attr_stdcall:
> +  case attr_thiscall:
> +  case attr_pascal:
> +  case attr_pnaclcall:
> +  case attr_inteloclbicc:
> 
> This seems to be missing the pcs attributes.  If that's intentional, there should be a comment.
> 
> Oops.  I'll make this a fully covered switch.
>  
> -  if (handleFunctionTypeAttr(state, attr, declSpecType)) {
> -    spliceAttrOutOfList(attr, attrList);
> -    return true;
> -  }
> -
> -  return false;
> +  return handleFunctionTypeAttr(state, attr, declSpecType);
> 
> This is an unexplained change.  Doesn't this cause the attribute
> to be potentially applied twice?
> 
> I had to apply this change in order to get one of the TypeLocBuilders to succeed without asserting in fillAttributedType().  What might happen if the parsed attribute stays on the declspec?

Oh, I see.  You're right, there's no reason to splice this out.

John.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20130719/32af1070/attachment.html>