2019-08-01 [18:00:40.0000] devsnek: at long last!! https://bugs.webkit.org/show_bug.cgi?id=200199 [18:00:48.0000] woo [18:01:11.0000] I didn't expect that the redo would take me a week lol [18:01:15.0000] so you went with a stack [18:01:20.0000] yep [18:01:32.0000] very nice [18:01:45.0000] I had it working with passing the label down, but I realized I created a ton of cruft in the process [18:01:51.0000] its kinda scary how much stuff you have to change though [18:01:56.0000] like instead of the patch doubling, it had tripled [18:02:37.0000] haha yeah [18:03:00.0000] in v8 i just emit the two JumpIfs and it magically does it all from there [18:03:48.0000] I actually sunk a day by making an seemingly-innocuous refactor that 100% broke JIT in debug [18:04:08.0000] like I would hit an debug assert just by checking 1+2 === 3 in a loop lolol [18:04:09.0000] such is life :) [18:05:00.0000] yeah I'm excited and relieved to reach the conclusion [18:05:44.0000] i wonder if sm has a patch yet [18:06:21.0000] hmm doesn't look like it does https://bugzilla.mozilla.org/show_bug.cgi?id=1566143 [18:07:29.0000] I'll make a GH thread [18:14:20.0000] https://github.com/tc39/proposal-optional-chaining/issues/115 [18:14:58.0000] i don't see anything from chakra yet [18:21:08.0000] also wow your impl is so different [18:21:22.0000] and by that I mean wow V8's bytecode generator is so different [18:22:24.0000] diverse implementations :) [18:22:30.0000] love looking at how different engines do things [18:23:15.0000] indeed [18:23:34.0000] if they're all the same it's kind of just a monoculture in disguise :P [11:11:42.0000] apologies if this is considered to be in poor taste, but: my (small) team at Shape Security (which is on tc39) is hiring people with esoteric web platform knowledge: https://twitter.com/disnet/status/1156982795501092864 [14:42:10.0000] /me takes a drag off the File API [14:51:08.0000] bradleymeck: careful, this ain't the APIs we grew up with [15:08:30.0000] devsnek: oh wow I just realized if I collect all the labels and emit them at once at the very end then that solves the single-use problem [15:10:17.0000] oh I thought you were already doing that with the stack lol [15:12:54.0000] haha [15:13:27.0000] I hadn't actually investigated that part [15:13:49.0000] incidentally I tried seeing what happens when you reuse a label and it produced a lovely infinite loop [15:13:50.0000] lol 2019-08-02 [14:24:10.0000] I was going to ask for a status update on private fields in TypeScript but I see the PR was reviewed 12 minutes ago. https://github.com/microsoft/TypeScript/pull/30829 2019-08-05 [20:40:06.0000] devsnek: wowsers, apparently I still wasn't doing `a?.b?.()` (combined optional access and optional call) correctly [20:40:26.0000] what was it doing [20:42:13.0000] I had it parsing as FunctionCallValueNode > OptionalChainNode > DotAccessorNode, but that runs into the same "loss of this" problem that I originally faced for `a.b?.()` [20:42:40.0000] i.e. for `a = { b() { return this.c; }, c: 42 }`, `a?.b?.()` would be undefined [20:43:25.0000] not sure how JSC-specific this problem is [20:44:24.0000] it's not hard to fix but I have to basically introduce an `OptionalChainType::CombinedAccessAndCall` case where I reach for the *second* label on the stack :( [20:45:06.0000] rkirsling are you implementing `?.` in JSC? =D [20:45:42.0000] drousso: https://bugs.webkit.org/show_bug.cgi?id=200199 :D [20:46:24.0000] the two places Darin requested extra coverage both had edge-case bugs lol [20:46:35.0000] best kind of review :chefkiss: [20:46:44.0000] aaaawwwwweeeeessssooooommmmmeeeee [20:46:53.0000] i cant wait to use that in Web Inspector =D [20:47:26.0000] :D :D :D [20:51:25.0000] rkirsling: every time you talk about jsc internals it scares me more and more [20:52:20.0000] lol it's really just the function node menagerie [20:52:59.0000] what is `a.b()` in V8? [20:55:04.0000] Call { expression = Property { base = ..., name = "b" } } [20:59:33.0000] ahh yeah it would sure be nice to have "function call" be a singular concept from the perspective of the bytecode generator [20:59:35.0000] hehehe [20:59:56.0000] in the bytecode generator it just branches based on what `expression` is [21:00:18.0000] and then you end up with a `callee` register and a `receiver` register [21:00:50.0000] i can't even find the parser in spidermonkey [21:07:41.0000] lol it's https://searchfox.org/mozilla-central/source/js/src/frontend/Parser.cpp and vicinity [21:14:36.0000] oh i thought the frontend folder would have like [21:14:40.0000] devtools or smth [21:15:28.0000] haha yeah it's quitehard to find the first time [07:54:45.0000] oh hey. qq. Has there are been a proposal or suggestion for something along the lines of array.protoype.difference? [08:05:51.0000] MylesBorins: comes up on esdiscuss everyone once in a while [08:11:01.0000] devsnek do you have a link? [08:11:57.0000] someone from the node project was pinging me about wanting to push forward a proposal for it... but wasn't able to find prior discussion [08:14:08.0000] I could find them in a bit [08:14:16.0000] iirc esdiscuss.org is searchable [08:14:47.0000] I wish we could translate old threads into discourse [08:41:21.0000] MylesBorins: What's the semantics of an array difference? Diffing has a lot of possible impls... [08:44:31.0000] TabAtkins what was pitched to me is a function that would return items that don't exist in both arrays [08:44:37.0000] e.g. [1, 'a', 3.14].difference([2, 'a', 3.145]) // returns [1, 3.14] [08:44:44.0000] [1, null].difference([2, null]) // returns [1] [08:44:45.0000] So that's a Set difference [08:45:07.0000] slightly different [08:45:10.0000] as per example 2 [08:45:21.0000] Ah, that's a Set minus. [08:47:38.0000] if those set methods happen, engines can probably optimize new Set(x).difference and new Set(x).minus [11:16:03.0000] MylesBorins: it seems like you'd almost always just want to be using sets in that case, and converting to a set and back would probably be clearer [11:16:44.0000] Bakkot so would it then be a good idea to explore making those set methods? [11:16:58.0000] MylesBorins: https://github.com/tc39/proposal-set-methods [14:49:01.0000] rkirsling (or anyone else who works on JSC): are you in a position to tell me if the rdar link at the botton of https://bugs.webkit.org/show_bug.cgi?id=38970 has had any progress? [14:49:09.0000] just for my own curiousity mostly [14:49:42.0000] unfortunately as a non-Appler, I have no idea what's on the other side of that protocol :( [14:50:23.0000] that's where they keep the apples [14:50:48.0000] alas [14:51:01.0000] devsnek: 👏 [14:52:13.0000] I always assume that any discussion of consequence would appear publicly but I could be wrong [14:53:44.0000] maybe drousso could confirm :p [15:01:14.0000] ljharb: gonna bring globalThis for stage 4 sometime? [15:38:51.0000] also: is there anyone around who remembers why `super` works the way it does? https://github.com/tc39/ecma262/issues/1654 [15:43:09.0000] ouch. [16:37:22.0000] Bakkot: when I’m ready for everyone to yell at me on twitter and github again [16:37:30.0000] hah, fair enough [16:37:42.0000] i've still got that that pr to fix extending null :( [16:40:07.0000] devsnek: is that ready for review again? if it's good to go, and I think it looks right, I can bring it before the committee in October [16:40:21.0000] i don't remember anymore [16:40:24.0000] (https://github.com/tc39/ecma262/pull/1321, for everyone else) [16:40:48.0000] oh right the argument about how super is allowed 2019-08-06 [17:16:58.0000] devsnek: it seems important to fix, if you and Bakkot can agree on what should go to the committee :-) [02:40:07.0000] /me trying to find some pattern -> https://twitter.com/GNUmanth/status/1158673972801593344 [07:26:33.0000] Is there a PDF, or downloadable (offline) version of the specification? [07:45:08.0000] Someone should add a service worker [07:54:51.0000] I would love that, happy to add! [07:57:18.0000] lol wouldn’t that make it slower tho [08:01:45.0000] Why would it be slower? It’s just a cache [08:01:56.0000] And only kicks in when you’re offline [08:03:21.0000] airbnb tried service workers but their mere presence on the page had large enough perf consequences that we yanked them out [08:04:37.0000] :( do you know what exactly was causing the problem? They shouldn’t be slowing down the browser [08:06:00.0000] I can be on trains/planes sometimes so it’s good to be able to read the spec without needing a constant internet connection. If there’s a way for me to try it without impacting the site I’m happy to spend time on it [08:06:29.0000] nope, but google’s sw team worked with the airbnb engs on it and couldn’t figure out how to solve it. [08:06:59.0000] the spec is already offline tho, once you load it once - i just keep a tab always open and it’s offline for me [08:07:45.0000] Haha that does not work well on tablets, on a pc sure but the page is so big my iPad constantly kills it and forces me to reload the page. [08:08:14.0000] Keeping the tab open isn’t an option for me [08:08:27.0000] ah, true [08:11:10.0000] Is it possible to experiment with a new path on the TC39 website that does this? Then we can compare the 2? I don’t know how the site is hosted currently but I remember it’s statically generated [08:31:04.0000] couldn't we auto-generate a PDF after auto-rendering the HTML, as part of CI? [08:32:52.0000] we already have stuff to generate PDFs for ecma [08:38:14.0000] if by “stuff” you mean [08:38:22.0000] “jordan prints to pdf on his Mac” [08:39:07.0000] after a year of manually fixing styles i at least tweaked the print stylesheet so i didn’t have to do it every time [08:40:45.0000] oh lol, i thought we had a script [08:42:44.0000] we do, it’s just written in dna and takes a very inefficient amount of energy to run :-p [08:48:08.0000] hard to invoke dna from ci [08:53:24.0000] I've raised https://github.com/tc39/tc39.github.io/issues/166 [08:57:10.0000] we used to have a sw on the nodejs docs [08:57:12.0000] but it got removed [08:57:23.0000] i wonder if it was for the same reason [09:01:06.0000] JaseW: shouldn't that go to ecma262? the website repo is literally just the website [09:01:30.0000] an org admin can move the issue [09:01:36.0000] bterlson: akirose: ^ [09:01:48.0000] oh nice I didn't know that feature existed! [09:01:56.0000] it's fairly recent [09:02:09.0000] i use it like once a day to move issues from nodejs/node to nodejs/help lul [09:02:40.0000] @rkirsling oh ok nice, wasn't aware of that [09:03:04.0000] good idea regardless though :D [09:03:19.0000] I've moved it https://github.com/tc39/ecma262/issues/1656 [09:03:42.0000] oh i guess delegates can do that too [09:03:48.0000] ljharb: do you still want me to move that css into a separate file [09:13:36.0000] yeah that’d be nice [10:40:36.0000] ljharb: I have been meaning to see if I can get ecmarkup to generate LaTeX. It's probably doable, yeah? [10:43:15.0000] Bakkot: that's much more bterlson's domain than mine, and i'm not that familiar with authoring latex (only reading all my calculus assignments made with it), but it seems like it should be doable [10:43:26.0000] bterlson ^ [10:56:01.0000] seems like it'd just be a matter of doing it [10:57:33.0000] a lot depends on the extent to which the spec contains raw HTML or (especially) CSS, and the extent to which ecmarkup's auto-linking facility is tied to the HTML representation [11:09:08.0000] Why do people keep insisting on these legacy formats... [11:09:13.0000] Use web pages. [11:10:16.0000] web pages are way more ephemeral and always will be [11:10:47.0000] cf fraction of wikipedia citations to urls which are now untraceable vs fraction of wikipedia citations to books which are now untraceable [11:19:31.0000] always cite via web.archive.org [14:54:20.0000] devsnek: wow, turns out adding a "jump if undefined or null" op to JSC was really easy [14:54:27.0000] nice [14:54:34.0000] after my initial CL lands i'm planning to add one to v8 [14:54:56.0000] yeah I can't really submit it until my existing patch lands [14:55:21.0000] seems like the Applers are busy of late [15:01:58.0000] rkirsling / devsnek: would strongly recommend adding a comment explaining why that operation is not suitable for `== null`; I know V8 has had at least one bug where someone assumed those were equivalent [15:02:16.0000] a comment to what [15:02:35.0000] oh the op [15:02:37.0000] hehe [15:02:39.0000] Bakkot: Yusuke had already added an OpIsUndefinedOrNull to address an IsHTMLDDA bug in ArraySpeciesCreate [15:03:15.0000] I'm just fusing it to `jtrue` / `jfalse` [15:03:56.0000] there's another fixme in here that leads me to believe I'll have also fixed a RequireObjectCoercible bug by doing this, but I'll find out shortly :D [15:07:44.0000] that said a comment might be good but I'm not quite sure where I'd put it; if they look up the definition, they'll see it's quite different, but yeah 🤔 [15:22:03.0000] hmm, guess there's no test262 case for something like `let { 0: x } = document.all;` [15:23:54.0000] but I did just fix that, heh. [16:01:49.0000] lol devsnek (/ Bakkot) it turns out the patch didn't technically conflict so I just submitted it separately: https://bugs.webkit.org/show_bug.cgi?id=200480 [16:02:37.0000] jsc scares me [16:03:04.0000] jsc's optimization is pretty rad. [16:10:22.0000] i like v8's layers [16:10:23.0000] fascinatingly the various "jump if " ops are just a bytecode thing [16:10:27.0000] like an onion [16:11:39.0000] so they apply for direct interpretation and baseline JIT, but for the higher stages of JIT, there's just a single "Branch" node in the dataflow graph [16:12:44.0000] so I didn't have to add anything there [16:14:15.0000] devsnek: how do you mean layers? [16:14:47.0000] devsnek: you were never involved in the crankshaft days, yeah? [16:14:59.0000] there used to be like half a dozen different backends [16:14:59.0000] it was so painful [16:15:04.0000] turbofan + ignition is much nicer [16:15:05.0000] i started using js around the time crankshaft was removed [16:15:23.0000] i only started contributing to v8 like last year though [16:15:58.0000] rkirsling: like in order to add a new opcode in v8, you build on primitives that may or may not translate into machine code [16:16:29.0000] ah hmm [16:16:33.0000] JumpIfUndefinedOrNull can be two WordEqual() calls which then get handled by each architecture's backend [16:16:55.0000] interesting [16:16:57.0000] actually i could probably use a bitmask [16:17:24.0000] I mean to be fair I'm not having to thing about individual architectures there [16:17:48.0000] that's a whole separate, and to me far scarier, area [16:17:53.0000] or series of areas even [16:18:07.0000] yea the only time i've ever touched that part was to delete stuff :P [16:18:08.0000] *think about [16:18:33.0000] I had to fix Win64 JIT at one point [16:18:55.0000] it took weeks and I think I would've been screwed entirely had a coworker not helped me out [16:19:05.0000] i still have no idea how a lot of v8's jit works [16:19:20.0000] learned a thing or two about calling conventions though 🤷 [16:19:38.0000] oh yeah had to dig into that for a wasm change i'm working on [16:19:46.0000] optimized way to call C from wasm [16:29:16.0000] but yeah the undefined or null check there is based on the fact that undefined is 0b1010 and null is 0b0010, so you just clear the undefined bit and see whether you're left with a null value [16:31:29.0000] it's eqnull that bakes in all the document.all checks -- my understanding is that they get optimized away, but still a burden for the interpreter 2019-08-07 [17:53:50.0000] aw man, V8 decided to call ?? just "the nullish operator", huh [17:53:54.0000] this is unfortunate [17:54:49.0000] ooo [17:56:13.0000] what a weird implementation [22:13:54.0000] wow devsnek your design doc is so thorough o_o [22:14:05.0000] what is `Star` anyway? get value? [22:14:25.0000] store accumulator in register [22:14:42.0000] and Lda is load accumulator [22:14:53.0000] right I figured it must be store in register right after typing that question [22:15:08.0000] i'm glad someone likes my doc [22:15:09.0000] couldn't remember what A was in either case though 😅 [22:15:17.0000] i thought it was rather skimpy [22:16:14.0000] if it were me I would've just linked the README and called it a day lol [22:16:50.0000] v8 requires design docs for everything [22:16:53.0000] (I mean in addition to typing out a general description like I did in the WK changelog) [22:17:07.0000] often helpful, sometimes superfluous, like in this case [22:17:08.0000] guess I didn't realize [22:18:08.0000] one interesting thing is how the bytecode generator reorders stores and loads [22:19:06.0000] jumps after stores [22:19:13.0000] i assume thats for cache locality or something [22:20:03.0000] hm hm [22:20:27.0000] but yeah in your case all the more you'd want that new jump op [22:22:20.0000] is this a thing https://gc.gy/32860336.png [22:22:29.0000] or did someone just make an account named anonymous nyan cat [22:22:34.0000] lol whaaa [22:22:56.0000] oh no its actually a thing [22:23:01.0000] I think it's me though? which means it's a thing [22:23:14.0000] I just happened to open it in incognito for no particular reason [22:23:56.0000] apparently there's also an anonymous kraken, pumpkin, and ifrit [22:24:12.0000] `An ifrit is an enormous winged creature of fire,` [22:24:37.0000] and more importantly a perennial FF summon [07:15:25.0000] I'm using entirely too many super override classes for private fields [08:00:20.0000] mmm trying to figure out if this is a v8 bug or proper spec... if I have a super override to init private fields, it seems like calling super(o); twice will initialize the private fields twice and thus throw errors, so the fix is to try/catch before calling super and return if o already has them installed? [08:02:28.0000] calling super twice sounds dangerous [08:03:48.0000] devsnek: its a hell of a drug [08:04:06.0000] also only real way to do private fields in a sharable way for now :-/ [08:04:25.0000] moving off weakmaps and seeing speed gains :choo choo: [08:05:03.0000] latest chrome seems to differ from node on this [08:07:51.0000] https://codepen.io/bradleymeck/pen/GVQaNv?editors=0012 works, but is... ugly? [08:10:15.0000] that try/catch makes me so irrationally angry [08:10:31.0000] the hasInstance one I mean [08:10:54.0000] haha [08:11:03.0000] i'd love `has` [08:11:39.0000] https://www.npmjs.com/package/@snek/private [08:11:45.0000] this is what I've been using [08:11:48.0000] no thanks ;p [08:12:12.0000] this works pretty well even if it looks gross it means I don't have a C++ dep [08:12:31.0000] which is something i'd like to keep for distribution issues [08:12:53.0000] yeah probably not ideal for libraries [09:27:34.0000] leobalter: very exciting to see TLA tests 👍🏻 [09:27:50.0000] nice :) there are more to come today [09:49:51.0000] leobalter: I think we also need tests for `09.1_1` and `09e1_1` being errors [09:51:09.0000] rkirsling would you open a PR for this, please? [09:51:22.0000] sure! [10:20:47.0000] devsnek: guess that answers my feature flag question 😎 [10:21:00.0000] hmm? [10:24:45.0000] oh that evidently it is flipped on in Safari TP [10:25:49.0000] nullish coalescing? [10:26:04.0000] yup! [10:50:38.0000] i really wish we had a way to consume a stream of JSON values [10:51:36.0000] i think whatwg had an async json or json streaming thing [10:55:18.0000] i know .json() theoretically allows off thread allocation to happen [10:57:16.0000] i wonder if for unknown values, parse off thread and copy is better than parse on main thread [10:57:32.0000] using serdes [10:58:57.0000] serdes would need to do the allocations into the vm heap still, probably won't beat existing stuff speed wise. also does serdes preserve key ordering?? [11:00:09.0000] if main thread is janking or has a queue for jobs its probably safe to start off thread since it would require a new tick to actually make the value available to user code [11:00:17.0000] serdes is how engines do structured cloning [11:00:31.0000] postMessage and such [11:00:49.0000] afaik they all just copy memory from one heap to another [11:01:29.0000] but thats full serialization, that removes the refs and hydrates them on other end for things like Array.prototype and Object.prototype? [11:02:20.0000] i don't see serdes in v8's default serializer at least [11:02:34.0000] i mean the intermediate bytes are basically just a memory copy [11:03:23.0000] in v8, they can also get away with not modifying prototypes and stuff because they all share root prototype maps [11:05:53.0000] would be nice [11:07:06.0000] i think they had a blog post about sharing intrinsics at some point [11:11:24.0000] thats the mem refs for the code of intrinsics, but the object refs on heap are still unique and allocated [11:12:14.0000] i just mean like, in every heap, the object for array's prototype has the same offset [11:12:27.0000] otherwise Object.freeze in one thread would affect other threads [11:12:35.0000] not that they share objects [11:12:39.0000] ah, yes [12:01:35.0000] devsnek WDYT? https://github.com/tc39/test262/pull/2274/commits/31d9a7ae2b28971876597fbc4f2adbab6f556fa6 [12:03:45.0000] devsnek nevermind, some stuff to fix in there :P [12:06:51.0000] leobalter: if you want, somewhere in engine262's commit history there should be an implementation of TLA [12:06:58.0000] might be on the experimental branch [12:07:23.0000] I removed it back in December so the memory is fuzzy [12:11:12.0000] devsnek this one is correct... https://github.com/tc39/test262/pull/2274/commits/db51745742bf91eb09df8bd7bf7954388b2d4568 [12:11:26.0000] devsnek hopefully it's back soon :) [13:40:24.0000] MylesBorins devsnek rkirsling this is ready for reviews if anyone is interested. https://github.com/tc39/test262/pull/2274 [13:41:43.0000] the deeply-nested ones made me laugh 😂 [13:52:59.0000] mfw 200 files [13:53:22.0000] https://gc.gy/32916200.png [14:20:29.0000] well, there is a lot of tests for syntax and we end up with files matching each production [14:21:08.0000] the one on the pic: lots of Blocks and an Object literal at the end :P [14:28:43.0000] https://github.com/tc39/test262/pull/2274/files#diff-c11ccae7ea60da1e6dc9af0edb170779 [15:07:29.0000] devsnek: grats sir [15:07:42.0000] 👍🏻 [15:07:50.0000] I'm a little sad you beat me to the punch :p [15:08:07.0000] well, i don't know if mine passes test262 yet [15:08:17.0000] i have to do that in a separate change [15:09:30.0000] ∀ test ∈ ∅ ... lol [15:10:19.0000] all tests that are a member of the empty set? [15:11:37.0000] yeah I mean it'll pass test262 so long as there's nothing to check :p [15:11:45.0000] oh haha [15:11:50.0000] oh wait a few were merged weren't they [15:11:56.0000] yeah two days ago [15:12:37.0000] but apparently if i want to update test262, i have to also triage *every failure* that happens as a result of it [15:12:58.0000] you sure do! :p [15:13:04.0000] so i figured i'd leave that to someone with a bit more comfort in v8 [15:13:19.0000] yeah I lucked out with my early ref error abolishment patch [15:13:43.0000] in that if I updated juuuust to the sha hash I needed then I didn't incur any new irrelevant failures [15:59:17.0000] chicoxyzzy: looking into how a feature flag can be flipped [15:59:31.0000] I too was misled by the announcement :sad_panda: 2019-08-08 [20:11:21.0000] ljharb? [20:11:59.0000] hi [20:12:58.0000] In 1376 you changed lots of old-style intrinsics to new-style, but didn't change all. [20:13:38.0000] I was wondering why not all, what the criterion was. [20:14:29.0000] jmdyck: i changed all the ones that were accessible off of other ones - which did i miss? [20:15:05.0000] jmdyck: like, `%ArrayPrototype%` is `%Array.prototype` because `%Array%` exists, but i didn't change things like IteratorPrototype etc [20:15:14.0000] if there's more that can be changed to use the new style, that'd be great [20:15:34.0000] I found about 67 [20:16:48.0000] E.g., a bunch of %GeneratorPrototype% that could be %Generator.prototype% [20:17:18.0000] basically i was only thinking about the user-observable ones [20:17:29.0000] but yeah you're right, those should be changed too [20:17:35.0000] if you make a PR, great, otherwise i'll do it tomorrow :-) [20:18:03.0000] yeah, I've got the search-and-replace commands set up. [20:18:13.0000] sweet, appreciate it [20:18:37.0000] just need to be sure to keep the old style ones for layered specs; i'll remove all of those once i've updated html and intl to use the new style. [20:19:08.0000] There are some where the new-style is a bit round-about, e.g. %AsyncGenerator% -> %AsyncGeneratorFunction.prototype%, you still want those to change? [20:19:11.0000] hey while you two are here -- is https://github.com/tc39/ecma262/pull/1658 actually meant to be normative? I thought it was just making explicit what was already assumed [20:19:38.0000] (oops, come back to me if I'm interrupting) [20:20:09.0000] jmdyck: its about making them more readable, so im fine with your judgement call there [20:20:30.0000] rkirsling: i’m not really sure [20:21:00.0000] rkirsling: if it changes observable requirements, even if it’s a spec bug that nobody we know about has implemented spiritually wrong, it’s still normative [20:21:13.0000] re old style for layered specs: sure, they're all in table 7, I'm not touching that. [20:21:13.0000] right that's fair [20:25:31.0000] ljharb: turns out %AsyncGenerator% was the only such example. I'll leave it as is. [20:29:43.0000] sgtm [20:34:30.0000] ljharb: I also found some things that shouldn't have been changed, so have commits to change them back. [20:34:42.0000] O.o like what? [20:35:40.0000] %AsyncIteratorPrototype% -> %AsyncIterator.prototype% [20:36:02.0000] latter assumes %AsyncIterator%, which doesn't exist [20:36:08.0000] oops, thanks then [20:36:52.0000] (Maybe it should exist, I don't know.) [20:37:14.0000] the iterator helper proposal can address that :-) [20:41:26.0000] ^ iterator methods poposal says it exists [20:54:20.0000] Dang. [20:56:29.0000] E.g., if I change "This function is the %Promise_all% intrinsic object." to new-style, then the %Promise_all% in Table 7 won't get auto-linked. [21:21:54.0000] right, that’s why i left those [21:23:50.0000] Although it's weird, some get linked correctly anyway. [21:24:38.0000] E.g. FunctionPrototype and ObjectPrototype in table 7 link to the right place, even though the has the new-style reference. [21:31:00.0000] So if I keep the s to old-style, then the old-style refs (in table 7) get auto-linked, but new-style references (anywhere) don't. [21:33:56.0000] Given that trade-off, I think I'd vote the other way. [21:35:58.0000] i.e., let the old-style refs in Table 7 go unlinked. (The new-style ref in column 3 will get linked, so there's still a link on the row.) [21:37:15.0000] Note that this is only talking about cases where old-style and new-style differ, not the 'base' ones like %Array%' and '%String%'. [23:05:31.0000] devsnek: does Chrome Canary give you a way to pass your V8 feature flag, btw? [01:37:43.0000] rkirsling, yeah, --js-flags="" [08:57:43.0000] Does the language have any built-in setters [08:58:48.0000] Domenic: __proto__ [08:58:57.0000] lol --_-- [08:59:26.0000] Domenic: also function.prototype.caller and function.prototype.arguments [08:59:41.0000] oh and arguments.callee [08:59:54.0000] oh and all the properties of a mapped arguments object [09:00:06.0000] i think that's everything [09:00:12.0000] so its basically all legacy stuff [09:04:07.0000] Yeah [09:24:12.0000] gsathya: cool! seems like Safari TP might not have an equivalent thing 😥 [09:29:51.0000] does nullish coalescing have the restriction requiring parens [09:39:46.0000] Don't forget RegExp friends $0-$9 [09:40:03.0000] RegExp in general... [09:41:35.0000] devsnek: yes [09:42:05.0000] bradleymeck: is that in the spec? [09:42:10.0000] like i know it exist [09:42:12.0000] exists* [09:42:15.0000] but i can't find it [09:44:01.0000] whats this channel about [09:44:11.0000] I know tc39 [09:46:44.0000] devsnek: it doesn't seem so, but it probably should be listed XD [09:47:07.0000] satoriprints: we discuss javascript here [09:49:49.0000] devsnek: yeah mixing is prevented from being a valid production [09:50:40.0000] see, say, https://github.com/tc39/proposal-nullish-coalescing/issues/44 [09:51:49.0000] rkirsling: does the grammar in the spec restrict it? [09:52:04.0000] i was wondering after referring to the actual text [09:52:16.0000] maybe i'm just bad at reading spec text [09:52:46.0000] yeah, notice where it's BitwiseOR [09:53:01.0000] those are blocking logical ops [09:53:18.0000] oh i see [09:53:28.0000] it can go to LogicalOR or Coalesce [09:53:37.0000] it all makes sense now [09:54:23.0000] mhm. was a really nice simplification 😃 [09:57:24.0000] one nitpick I'd been meaning to make is that LogicalExpression is an unfortunate name since "logical" means "truth(iness)-oriented", though the only countersuggestion I can come up with is like ShortCircuitExpression [09:57:55.0000] well logicaland and logicalor are truthiness oriented [10:03:58.0000] but coalescing is nullish-oriented, so unless we mean ternary logic... 😛 short circuiting is really the thing that they have in common [10:07:52.0000] I phrased JSC's error message like this: https://github.com/WebKit/webkit/blob/master/Source/JavaScriptCore/parser/Parser.cpp#L3932 [10:31:32.0000] does jsc integrate with its own debugger via CLI or just lldb? [10:47:43.0000] bradleymeck: how do you mean CLI? lldb would be the command-line approach but you can use Xcode or Visual Studio [10:52:41.0000] rkirsling: d8 for example can be debugged using chrome devtools if you are shifty, XS has XSBug [10:52:49.0000] gui devtools basically [10:55:31.0000] oh yeah you can inspect a JSContext [10:56:54.0000] you do `run-jsc --remote-debug`, open Safari, and choose Develop >