[clang] [PAC] Add support for __ptrauth type qualifier (PR #100830)

[Aaron Ballman via cfe-commits]

================
@@ -0,0 +1,21 @@
+// RUN: %clang_cc1 -triple arm64-apple-ios -fsyntax-only -verify -fptrauth-intrinsics -std=c++11 %s
+// RUN: %clang_cc1 -triple aarch64-linux-gnu -fsyntax-only -verify -fptrauth-intrinsics -std=c++11 %s
+
+template <typename T> struct G {
+  T __ptrauth(0,0,1234) test;
+  // expected-error at -1 2 {{type '__ptrauth(0,0,1234) T' is already __ptrauth-qualified}}
+};
+
+template <typename T> struct Indirect {
+  G<T> layers;
+  // expected-note at -1{{in instantiation of template class 'G<void *__ptrauth(0,0,1235)>' requested here}}
+  // expected-note at -2{{in instantiation of template class 'G<void *__ptrauth(0,0,1234)>' requested here}}
+};
+
+void f3() {
+  Indirect<void* __ptrauth(0,0,1234)> one;
+  // expected-note at -1{{in instantiation of template class 'Indirect<void *__ptrauth(0,0,1234)>' requested here}}
+  Indirect<void* __ptrauth(0,0,1235)> two;
+  // expected-note at -1{{in instantiation of template class 'Indirect<void *__ptrauth(0,0,1235)>' requested here}}
+  Indirect<void*> three;
+}
----------------
AaronBallman wrote:

> @AaronBallman
> 
> __ptrauth qualified pointers can be template parameters, and can be deduced, however the qualifier itself cannot be dependent, e.g.
> 
> ```c++
> template <class T> struct Foo {
>   T v;
>   T *__ptrauth(1,1,1000) ptr;
> };
> ```
> 
> Is valid, and can be instantiated as you would expect with `Foo<int>`, `Foo<int *>`, or `Foo<int *__ptrauth(1,1,1)>`.

With `Foo<int *__ptrauth(1,1,1)>` you would get an error because of the mismatch?

> However you can't use template parameters in the qualifier itself, e.g.
> 
> ```c++
> template <unsigned u> struct Foo {
>    void * __ptrauth(1,1,u) ptr;
> }
> ```
> 
> is not valid, but that's just because supporting such has never been high enough on the priority list.

I think it would make sense to support that, but I don't think it's required for the initial implementation. I'd say we should add a test case to show the current behavior and maybe file an issue so we don't forget the improvement.

https://github.com/llvm/llvm-project/pull/100830