Fix possible memory leak and remove superfluous NULL checks before free()

[anmolsarma]

No description provided.

[hawicz]

This change is either incomplete, or wrong. If setlocale(LC_NUMERIC, NULL) can never return NULL, then the use of tmplocale is superfluous and can be removed. On the other hand, if that can return NULL, then it's possible for oldlocale to be NULL, so the NULL checks are required.

[anmolsarma]

@hawicz If the pointer passed to free() is NULL, no action occurs. It is not necessary to check if a pointer is NULL before calling free() on it.

[hawicz]

Sory, but json-c is used on many different systems and on some it is necessary to check for null. Just because you can get away with it on your particular OS with your particular malloc implementation is not a reason to break it elsewhere.

[rgerhards]

@hawicz just out of curiosity: which currently existing system is so broken that it is necessary to check for NULL when calling free? It's long standardized that this is a no-op, see http://www.open-std.org/JTC1/SC22/wg14/www/docs/n1124.pdf section 7.20.3.2.

[hawicz]

hmm... it seems that I'm wrong about calling free(NULL) being a problem. It still feels wrong to me, but perhaps that just me getting old and crufty. In some ways I like the fact that a check for null helps imply what's values are expected to be present, but now not enough to block this change.