[llvm] [X86] Reject 'p' constraint without 'a' modifier in inline asm (PR #185799)

[Fangrui Song via llvm-commits]

================
@@ -7,7 +7,7 @@ define ptr @foo(ptr %Ptr) {
   %Ptr.addr = alloca ptr, align 8
   store ptr %Ptr, ptr %Ptr.addr, align 8
 ; CHECK: movq    %rdi, -8(%rsp)
-  %1 = tail call ptr asm "mov $1, $0\0A\09lea $2, $0", "=r,p,*m,~{dirflag},~{fpsr},~{flags}"(ptr %Ptr, ptr elementtype(ptr) %Ptr.addr)
+  %1 = tail call ptr asm "mov ${1:a}, $0\0A\09lea $2, $0", "=r,p,*m,~{dirflag},~{fpsr},~{flags}"(ptr %Ptr, ptr elementtype(ptr) %Ptr.addr)
----------------
MaskRay wrote:

@aemerson added this x86 test in https://reviews.llvm.org/D155023 but I think it's incorrect. In clang, `%1` dereferences the memory, unlike GCC.


https://github.com/llvm/llvm-project/pull/185799