00:35
<Michael Ficarra>
@bakkot explain this: https://github.com/tc39/ecmarkup/blob/main/src/Spec.ts#L1878-L1887
00:36
<bakkot>
lol
00:36
<bakkot>
I mean, that one is necessary because ecmarkup outputs utf8
00:37
<bakkot>
significantly predates me https://github.com/tc39/ecmarkup/commit/ce5fcdd77cda052fdadc93b0f6bf12f69ecf2f98
17:28
<ljharb>
before i merge the pending things, could perhaps https://github.com/tc39/ecma262/pull/3480 land so i don't have to manually check the IPR pre-merge anymore?
17:30
<Michael Ficarra>
I'm not interested in reviewing it, so if @bakkot says it's good, it's fine by me
17:30
<bakkot>
yeah ship it
17:36
<Michael Ficarra>
@bakkot wanna stamp https://github.com/tc39/ecma262/pull/3487 if you approve of that plan?
17:55
<bakkot>
wfm
18:11
<bakkot>
ljharb: https://github.com/tc39/ecma262/pull/2924 should land as one commit
18:11
<ljharb>
ah ok, will fix
18:11
<bakkot>
as is the second one is basically reverting the first; no reason to have both
19:19
<bakkot>
shu: https://github.com/tc39/ecma262/issues/3117 we can just pull the 1. Let type be TypedArrayElementType(typedArray). into the abstract closure, yeah? that way it won't happen until after the validation. and it's pure so that's allowed even inside of a read-modify-write modification function
19:20
<bakkot>
alternative is that we do the validation eagerly there and then allow it to be re-done in the AtomicReadModifyWrite call
19:20
<bakkot>
I guess I'm fine with either
19:52
<shu>
bakkot: yeah i'd just put it into the AC
19:53
<shu>
both the type and the little endian thing don't need to be closed over, actually
19:56
<bakkot>
depends a little on what you are allowed to do inside of a "mathematical function", which the read-modify-write functions are supposed to be
19:56
<bakkot>
but I'm fine with it
19:58
<shu>
spec-wise the requirements are: They perform all their algorithm steps atomically. Their individual algorithm steps are not observable.
19:59
<shu>
both getting the type and getting the endianess satisfy both counts. implementations should understand the as-if nature that getting the type can be lifted out
19:59
<shu>
(and since obviously no RMW operations are implemented like this with a higher order function)