Consider removing call to named property setter in legacy platform object [[Set]] #630
Comments
|
Hmm. It's possible this [[Set]] algorithm was written or changed after the Gecko implementation. I don't remember the exact history here. It's worth looking up the history of this part of the spec to see what the reasoning for it was, exactly. |
|
I think the last major change to the Gecko implementation was https://bugzilla.mozilla.org/show_bug.cgi?id=987007 (April 2014). The spec was written in #45 (March 2015). |
|
Ah, so the key question is how named creators (aka setters nowadays) should work. I think in Gecko if you do the thing on an [OverrideBuiltins] object (e.g. DOMStringMap) the behavior would be "foo", not "undefined"... |
|
Hmm, indeed. The test below doesn't call proto.[[Set]], and just calls the setter. Should I just add a check for <!DOCTYPE html>
<script>
var ds = document.documentElement.dataset;
var a = new Proxy(Object.getPrototypeOf(ds), {
set(...args) { w(args); return false; },
});
Object.setPrototypeOf(ds, a);
ds.x = "foo";
w(ds.x)
</script>
|
|
I confused myself again; it looks like WebKit is following the spec in all cases I tested. Results:
Testsno [OB], O = Receiver<!DOCTYPE html>
<script>
sessionStorage.clear()
var a = new Proxy(Object.getPrototypeOf(sessionStorage), {
set(...args) { w(args); return false; },
});
Object.setPrototypeOf(sessionStorage, a);
sessionStorage.x = "foo";
w(sessionStorage.x)
</script>no [OB], O ≠ Receiver<!DOCTYPE html>
<script>
sessionStorage.clear()
var a = new Proxy(Object.getPrototypeOf(sessionStorage), {
set(...args) { w(args); return false; },
});
Object.setPrototypeOf(sessionStorage, a);
Object.create(sessionStorage).x = "foo";
w(sessionStorage.x)
</script>[OB], O = Receiver<!DOCTYPE html>
<script>
var ds = document.documentElement.dataset;
var a = new Proxy(Object.getPrototypeOf(ds), {
set(...args) { w(args); return false; },
});
Object.setPrototypeOf(ds, a);
ds.x = "foo";
w(ds.x)
</script>[OB], O ≠ Receiver<!DOCTYPE html>
<script>
var ds = document.documentElement.dataset;
var a = new Proxy(Object.getPrototypeOf(ds), {
set(...args) { w(args); return false; },
});
Object.setPrototypeOf(ds, a);
Object.create(ds).x = "foo";
w(ds.x)
</script> |
… interfaces This is the case where it makes most sense, since the prototype chain can't shadow any properties. This change does not affect normal usage; it is only detectable by messing with the prototype of the object. The majority of implementations (Gecko and Chrome) already follow the proposed change; WebKit follows the existing spec. Fixes #630.
… interfaces This is the case where it makes most sense, since the prototype chain can't shadow any properties. This change does not affect normal usage; it is only detectable by messing with the prototype of the object. The majority of implementations (Gecko and Chrome) already follow the proposed change; WebKit follows the existing spec. Fixes #630.
Test:
Gecko and Chrome call proto.[[Set]]() rather than taking the early return, as shown by the log:
but WebKit may be following the spec:
The text was updated successfully, but these errors were encountered: