[clang] [Clang] Fix parsing of reversible type traits in template arguments (PR #95969)

[Aaron Ballman via cfe-commits]

================
@@ -141,3 +141,15 @@ namespace r360308_regression {
     return a == b;
   }
 }
+
+namespace GH95598 {
+template<typename _Tp, bool _IsPtr = __is_pointer(_Tp)>
+struct __is_pointer {};
+// expected-warning at -1 {{keyword '__is_pointer' will be made available as an identifier for the remainder of the translation unit}}
----------------
AaronBallman wrote:

Thanks for the historical information @zygoloid!

> I don't think we should be extending this hack further. I've suggested on the libstdc++ bug report that this be fixed in libstdc++ instead, by stopping using our keywords as their type names. Maybe one day we can then remove it entirely.

I tend to agree; if we can avoid continuing to extend this hack, I think we should. That said, I'd be curious whether deprecation is highly disruptive or not; my intuition is that we'd struggle with this one because people tend to use boost (and certainly libstdc++) as system headers, and we suppress diagnostics by default in system headers. So I'd expect this deprecation to take a *long* while before we could remove the hacks due to the long tail of not wanting to break system headers.

The ctor situation would be annoying, but I think the ctor and dtor are the only places we'd have to put in a special hack for "got a keyword where we expected an identifier", so perhaps the blast radius is sufficiently limited for that to be a reasonable approach?

https://github.com/llvm/llvm-project/pull/95969