gh-140544: fix build for including pycore_pystate.h when HAVE_THREAD_LOCAL is not defined

[kumaraditya303]

• Issue: Store current interpreter as a thread local variable #140544

[kumaraditya303]

cc @cdce8p

[cdce8p]

This fixes the issue. Thanks @kumaraditya303!

[ZeroIntensity]

LGTM, thanks.

[cdce8p]

Unfortunately it seems the last change broke it again.

/.../cpython/Include/internal/pycore_pystate.h:93:8: error: unknown type name '_Py_thread_local'
extern _Py_thread_local PyThreadState *_Py_tss_tstate;

/.../cpython/Include/internal/pycore_pystate.h:93:38: error: expected ';' after top level declarator
extern _Py_thread_local PyThreadState *_Py_tss_tstate;
                                     ^
                                     ;

It looks like _Py_thread_local only works for defined(HAVE_THREAD_LOCAL). So 34a2688 should probably be reverted.

cpython/Include/pyport.h

Lines 516 to 528 in f5394c2

	# define HAVE_THREAD_LOCAL 1
	# ifdef thread_local
	# define _Py_thread_local thread_local
	# elif __STDC_VERSION__ >= 201112L && !defined(__STDC_NO_THREADS__)
	# define _Py_thread_local _Thread_local
	# elif defined(_MSC_VER) /* AKA NT_THREADS */
	# define _Py_thread_local __declspec(thread)
	# elif defined(__GNUC__) /* includes clang */
	# define _Py_thread_local __thread
	# else
	// fall back to the PyThread_tss_*() API, or ignore.
	# undef HAVE_THREAD_LOCAL
	# endif

[ZeroIntensity]

No, this is all just HAVE_THREAD_LOCAL being messy. We should remove that all throughout the codebase and assume that thread locals are always available. mypyc will get some speedups from using the thread local directly. I'll put up a PR.

[cdce8p]

Sounds good. Happy to test it, just let me know.