| 08:06 | <ystartsev> | thanks bradleymeck & devsnek |
| 20:47 | <jorendorff> | Is this a spec bug? `Reflect.construct(Object, [], class C {})` seems like it'll trip an assertion in the spec |
| 20:48 | <ljharb> | jorendorff: which assertion? |
| 20:48 | <jorendorff> | because classes have [[Construct]], but IsCallable(C) is false |
| 20:48 | <ljharb> | no it's not |
| 20:48 | <ljharb> | (unfortunately) classes have a [[Call]] slot with an error-throwing function in it |
| 20:49 | <jorendorff> | oh, that makes sense |
| 20:49 | <jorendorff> | I was thinking of step 2 of https://tc39.es/ecma262/#sec-getprototypefromconstructor |
| 20:49 | <ljharb> | imo it doesn't make sense :-) it'd be much better if classes lacked a [[Call]] |
| 20:49 | <ljharb> | and we could still change that! but it'd require non-observable changes to `typeof`, and observable changes to Proxy |
| 20:49 | <ljharb> | and lots of non-observable spec changes |
| 20:50 | <jorendorff> | i knew that, just blanked on it |
| 20:50 | <devsnek> | some people have been known to use proxies to check for construct slots |
| 20:50 | <ljharb> | devsnek: the change i'd like to make wouldn't break that; it'd just let you *also* check for call slots. |
| 20:51 | <jorendorff> | ...which you can also do, but you'd get a different answer for classes |
| 20:51 | <devsnek> | when was [[Construct]] added |
| 20:51 | <ljharb> | jorendorff: true |
| 20:51 | <ljharb> | devsnek: ES6 |
| 20:51 | <devsnek> | sad |
| 20:52 | <devsnek> | should've just not had [[Construct]] imo |
| 20:52 | <ljharb> | that's also an alternative |
| 20:52 | <ljharb> | or was |
| 20:52 | <jorendorff> | alas :) |
| 20:53 | <devsnek> | its kind of interesting that all engines implement tail calls for Function.prototype.call and Reflect.call and such |