gh-157212: Fix _Py_ThreadId() on Windows ARM64 with MinGW

The __MINGW32__ && _M_ARM64 branch calls __getReg(18), mirroring the MSVC branch above it. The intrinsic is not available for MinGW ARM64: mingw-w64 declares __getReg only for Itanium, and GCC does not provide a corresponding builtin.

The branch is reached by AArch64 MinGW toolchains because mingw-w64 defines _M_ARM64 when __aarch64__ is defined.

Falling through to the generic __aarch64__ branch would not provide the Windows thread pointer. The Windows ARM64 ABI reserves x18 as the platform register pointing to the TEB, so read x18 directly.

Author: christianaurichzm

Include/cpython/object.h

@@ -523,7 +523,8 @@ _Py_ThreadId(void)
 #elif defined(__MINGW32__) && defined(_M_IX86)
     tid = __readfsdword(24);
 #elif defined(__MINGW32__) && defined(_M_ARM64)
-    tid = __getReg(18);
+    // x18 is the Windows ARM64 platform register and points to the TEB.
+    __asm__ ("mov %0, x18" : "=r" (tid));
 #elif defined(__i386__)
     __asm__("{movl %%gs:0, %0|mov %0, dword ptr gs:[0]}" : "=r" (tid));  // 32-bit always uses GS
 #elif defined(__MACH__) && defined(__x86_64__)

Misc/NEWS.d/next/Windows/2026-09-10-02-00-00.gh-issue-157212.x18TEB.rst

@@ -0,0 +1,3 @@
+Fix ``_Py_ThreadId()`` compilation on Windows ARM64 with MinGW toolchains,
+where ``__getReg()`` is unavailable. The TEB address is now read directly
+from ``x18``. Patch by Christian Aurich.