Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 98.38%. Comparing base (f6e23fd) to head (45aa65b).
Report is 2 commits behind head on master.

@@           Coverage Diff           @@
##           master   #17729   +/-   ##
=======================================
  Coverage   98.38%   98.38%           
=======================================
  Files         171      171           
  Lines       22238    22238           
=======================================
  Hits        21879    21879           
  Misses        359      359           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

Add test for ar.push.call(ar, 3) and ar.push.call(ar2, 3).

@ntoll @WebReflection FYI

I had a look and if the test works fine then I think it's OK if also these two work as expected #17729 (comment) which I believe is the case, as it shouldn't be too different from borrowing methods via the prototype ... however, I am also not super sure why there are so many invariants for the same thing but in general, at least in JS, I would use the apply trap for functions or something like __call__ trap in Python per each accessed reference, something to consider for the future?

In short, if you end up always with fn.call(js_context or js_undefined, ...args) a lot of code might be simplified but also with a proper trap there would be less guessing around and just natural consequences out of the meant code.

arrow functions and/or most global constructors would ignore the context, but in case of namespaces, context is ignored too, that is methods from Math or JSON or even window (or globalThis) so it feels like this dance might be reduced by far to me, but I let you decide if or when it'd be the right time to do so, also I need to find time myself to eventually contribute with a POC or something.

Again, if latest tests work, I think for the time being we're all good!

Thanks @WebReflection for the review.

I've now added those two extra tests and they pass.

so it feels like this dance might be reduced by far to me, but I let you decide if or when it'd be the right time to do so, also I need to find time myself to eventually contribute with a POC or something.

If the code can be simplified so it doesn't have to select between f(args) and f.call(args) that would be great, and I'd change it. But I just don't know how else to do it.

Anyway, the existing and new tests pass, so at least the semantics seem correct, even if they are not optimally implemented.