warn if an optional parameter is always passed

[michaelficarra]

This way we can mark it non-optional.

[ljharb]

Perhaps also we could warn if it's never passed, so we can remove it.

[jmdyck]

Perhaps also we could warn if it's never passed, so we can remove it.

The current ecma262 has a few cases where an operation's optional parameter is never passed an argument:

• CreateBuiltinFunction's _prefix_
• LoadRequestedModules's _hostDefined_
• DetachArrayBuffer's _key_ (but only because the spec never invokes the operation)
• GetModifySetValueInBuffer's _isLittleEndian_

but you probably don't want to remove the parameter in the first 3 cases, and maybe in the last one too.

[michaelficarra]

but you probably don't want to remove the parameter

I don't understand why we would want the parameter if it's never passed. We could always add it if we find a need for it.

[jmdyck]

You probably don't want to remove ...

• CreateBuiltinFunction's _prefix_ because the fact that it's never passed an argument is basically a spec bug. See Editorial: Specify _prefix_ for non-specific uses of CreateBuiltinFunction ecma262#2833.
• LoadRequestedModules's _hostDefined_ because, although the spec never passes an arg to it, HTML does.
• DetachArrayBuffer's _key_ because it's intended to be used by "certain embedding environments, not by algorithms in this specification".

As for GetModifySetValueInBuffer's _isLittleEndian_, I'm not sure. As far as I can tell, it'd be okay to remove, but I might be missing something.

[michaelficarra]

CreateBuiltinFunction's _prefix_

Fair.

LoadRequestedModules's _hostDefined_ and DetachArrayBuffer's _key_

If we mark integration points as such (tc39/ecma262#3147 (comment)), we can exclude these from the check.

GetModifySetValueInBuffer's _isLittleEndian_

Let's remove it: tc39/ecma262#3173