Tested impl: works! Thoudh... XXX: man says `FD_CLOEXEC` should be passable to `memfd_secret()`, but it fails with `EINVAL` if passed `FD_CLOEXEC`...? Find out why?

Fortune for memfd_secret-shim's current commit: Blessing − 吉

src/memfd_secret.c

@@ -12,7 +12,7 @@
 #define READ_ONCE(slot) ((__typeof__(slot))(*(const volatile __typeof__(slot)*)&(slot)))
 #define WRITE_ONCE(slot, value) (*((volatile __typeof__(slot)*)&(slot)) = (value))
 
-__attribute__((gnu_inline))
+__attribute__((gnu_inline/*, always_inline*/))
 static inline
 int _memfd_secret_raw(unsigned int flags)
 {
@@ -29,7 +29,7 @@ static inline
 int _has_memfd_secret_raw()
 {
 	// Attempt syscall
-	int fd = _memfd_secret_raw(FD_CLOEXEC);
+	int fd = _memfd_secret(0); //XXX: NOTE: man page says `FD_CLOEXEC` is a valid flag, but using it returns `EINVAL`?
 
 	// If failure to create new fd was caused by `ENOSYS`, it is not available.
 	if(fd < 0 && errno == ENOSYS)