[PATCH] D148702: [clang] Add Parse and Sema support for RegularKeyword attributes

[Richard Sandiford via Phabricator via cfe-commits]

rsandifo-arm marked an inline comment as done.
rsandifo-arm added inline comments.


================
Comment at: clang/lib/Sema/SemaDeclAttr.cpp:2902
     S.Diag(AL.getRange().getBegin(), diag::err_attribute_wrong_decl_type)
-        << AL << ExpectedTypeOrNamespace;
+        << AL << 0 << ExpectedTypeOrNamespace;
     return;
----------------
erichkeane wrote:
> Every where you are doing just a '0' in a diagnostic here it makes it incredibly unreadable.  I'd prefer 1 of 2 solutions:
> 
> 1- Create an enum somewhere that encodes the meaning here, and use those instead.
> 2- use a `/*Whatever*/` comment every place you're passing a raw literal.
Yeah, that's fair.  Sorry about that.  I got some of my own medicine reading the patch back after a while away from it.

Most of those `<< 0` come from looking at individual uses to see whether a regular keyword spelling is ever possible.  If it wasn't possible, `<< 0` was supposed to be a justification for not covering that line of code in the new test cases.

But I now think that was a mistake.  Using `isRegularKeywordAttribute()` is useful for readability even if we “know” what its value ahead of time.  And using it is more future-proof as well.

So this update converts most uses of `<< 0` to `<< Something.isRegulardKeywordAttribute` in cases where the appropriate `Something` is readily available.  There are a handful of cases where an immediate is still needed, so I went for option (2) and added a comment to those.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D148702/new/

https://reviews.llvm.org/D148702