1170742 - Update Promise.resolve to match latest spec (no [[PromiseConstructor]])

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[cscott]

TC39 agreed to get rid of the hidden [[PromiseConstructor]] property from the spec and simplify `Promise.resolve`. The adopted proposal also makes Promise.resolve ignore "species":

https://esdiscuss.org/topic/fixing-promise-resolve

Opening an issue to track adoption of the new semantics.

Steps to reproduce:

Execute this in the console:

p = Promise.resolve(5);
p.constructor = {}
p2 = Promise.resolve(p)
p===p2


Actual results:

"true"

Expected results:

According to latest spec, p===p2 should be "false".

Comment 1

[Boris Zbarsky [:bzbarsky]]

We don't really implement the old semantics, for what it's worth.  I'm guessing that we'll update some of this stuff as we implement Promise subclassing, though.

Comment 2

[cscott]

For what it's worth, you can implement the proper semantics now -- Promise.resolve() no longer depends on the details of subclassing, it uses a simple test of the "constructor" property.

As you can tell by the test case above, Firefox *does* currently implement the "hidden [[PromiseConstructor]] field" semantics of the old spec, which were errata'ed out of existence.  So it's worth changing in the near term to prevent this wrong behavior from spreading.

Comment 3

[Boris Zbarsky [:bzbarsky]]

Sure.  If the promise subclassing stuff doesn't turn out to be near-term (like a month or so), we can fix this up separately.  But I suspect it'll be near-term.

Comment 4

[Boris Zbarsky [:bzbarsky]]

Fixed in bug 1170760.