[PATCH] D75210: [Attr] Allow ParsedAttr to be constructor for any Attribute

Wed Feb 26 14:02:23 PST 2020

jdoerfert marked an inline comment as done.
jdoerfert added a comment.

In D75210#1894370 <https://reviews.llvm.org/D75210#1894370>, @erichkeane wrote:

> I don't see anything glaring here, but I would love some better explanation as to the goal here.  I don't see this used anywhere, the commit message is a bit underwhelming in its reasoning, and there is no test/etc showing why I would want this.

Agreed. The followup "user" patch is going to be put on phab very soon. I did rush this one out because I saw D31337 <https://reviews.llvm.org/D31337> land today. Wanted to make sure people see what I hope to achieve here.

The reason for this is `omp begin/end declare variant` which I implement via `attribute push/pop`. The idea of both are basically the same, attach some information (=Attribute) on all function (definitions) in the scope. I have a new function definition matcher and the rest of the logic basically done, some overloading work is still missing.

================
Comment at: clang/include/clang/Sema/ParsedAttr.h:115
 /// 3: __attribute__(( format(printf, 1, 2) )). ParmName/Args/NumArgs all used.
 /// 4: __attribute__(( aligned(16) )). ParmName is unused, Args/Num used.
 ///
----------------
erichkeane wrote:
> Can you update this comment?
Will do.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D75210/new/

https://reviews.llvm.org/D75210