Build failure - The Moirai have given mortals hearts that can endure.

Link to build: https://jenkins.wildfiregames.com/job/docker-differential/370/display/redirect

Tested on VS2015.

On exception handling:
https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JSNative if the function does not succeed:

it should either report an error (using e.g. JS_ReportError or JS_ReportOutOfMemory) or raise an exception (using JS_SetPendingException), and the callback must return false. (Returning false without reporting an error or raising an exception terminates the script with an uncatchable error.

This requirement is met, because GetPrivate calls JS_ReportError if it is not an object and returns false;
or if it is an object but could not obtain a pointer from JS_GetInstancePrivate, obtained nullptr.
The stack rooting and static_casts dont fail, or the error will segfault in a later statement.
JS_GetInstancePrivate does report an error in case of returning null:
https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JS_GetInstancePrivate

JS_PUBLIC_API void* JS_GetInstancePrivate(JSContext* cx, HandleObject obj,
                                          const JSClass* clasp,
                                          CallArgs* args) {
  if (!JS_InstanceOf(cx, obj, clasp, args)) {
    return nullptr;
  }
  return obj->as<NativeObject>().getPrivate();
}


JS_PUBLIC_API bool JS_InstanceOf(JSContext* cx, HandleObject obj,
                                 const JSClass* clasp, CallArgs* args) {
  AssertHeapIsIdle();
  CHECK_THREAD(cx);
#ifdef DEBUG
  if (args) {
    cx->check(obj);
    cx->check(args->thisv(), args->calleev());
  }
#endif
  if (!obj || obj->getJSClass() != clasp) {
    if (args) {
      ReportIncompatibleMethod(cx, *args, Valueify(clasp));
    }
    return false;
  }
  return true;
}

This code hasnt changed between sm45 and sm68.

From https://developer.mozilla.org/en-US/docs/Mozilla/Projects/SpiderMonkey/JSAPI_reference/JS_GetInstancePrivate

Note that if obj is an instance of clasp, but there is no private data currently associated with the object, or the object cannot have private data, JS_GetInstancePrivate also returns NULL.

So in this case, there should still be an error reported.
And it is reported in the JSNative that calls ScriptInterface::GetPrivate.

But the function should have a comment communicating this obligation to the caller. (Edit: The JSNative is the one making that requirement, i.e. none of GetPrivate's business)

(Also for reference, some brief discussions with wraitii on http://irclogs.wildfiregames.com/2019-08/2019-08-12-QuakeNet-%230ad-dev.log)

Further exception handling:

So the JSNatives that call this must still throw an exception in case the given JS Object is of the right class, but its data pointer being null.

Either the function supports nullpointer privates, then the caller has to JS_ReportError if it receives one.
But if it does that, then it overwrites the previous exceptions, making those useless.
And we dont want to add another output parameter I think.
...
Since all use cases assume that the private data is the C++ class instance that manages the JS object, and
since one can only distinguish the JS_GetInstancePrivate error case from the valid nullptr case by asking for JS_IsExceptionPending, it seems better to have the ScriptInterface::GetPrivate function in such a way that it always triggers an exception or returns the non-null class instance pointer.

To test this case where the object is of an appropriate class but has null prviate data, one can remove the JS_SetPrivate clal in IGUIObject.cpp.

Compiling takes forever.