| 20:55 | <rkirsling> | Bakkot: avp: apparently I read the spec too wishfully and it's altogether unspecified 😅 https://github.com/tc39/ecma262/issues/2039 |
| 20:55 | <rkirsling> | in my mind the expected behavior seems clear though; hopefully it's not just me |
| 21:12 | <Bakkot> | pop quiz: what does `/[\1]/.test('\u{1}')` evaluate to |
| 21:15 | <rkirsling> | oof |
| 21:16 | <rkirsling> | didn't realize `/[\1]/u` is a syntax error either |
| 21:16 | <Bakkot> | it isn't |
| 21:16 | <Bakkot> | it is not a syntax error |
| 21:18 | <rkirsling> | oh then that's a V8 bug whoops |
| 21:19 | <Bakkot> | my v8 does not give an error for that? |
| 21:21 | <rkirsling> | everybody but Ch gives a syntax error for me |
| 21:21 | <Bakkot> | uh |
| 21:21 | <Bakkot> | ... really? |
| 21:21 | <rkirsling> | not for your original example |
| 21:22 | <Bakkot> | oh! |
| 21:22 | <Bakkot> | I missed the `\u` |
| 21:22 | <Bakkot> | *`/u` |
| 21:22 | <Bakkot> | yes, with the `u` it's a syntax error |
| 21:23 | <Bakkot> | (because the only production which would allow it is `IdentityEscape: [~U] SourceCharacterIdentityEscape [?N]`, which is gated on `~U`) |
| 21:25 | <rkirsling> | yeah |
| 21:36 | <Bakkot> | sorry, rather, it would also be allowed by `CharacterEscape :: [~U] LegacyOctalEscapeSequence`, but that's also gated) |
| 21:57 | <rkirsling> | shu: what's your opinion on https://github.com/tc39/ecma262/issues/2039 |
| 21:58 | <rkirsling> | (namely whether it's worth trying to forbid \8 and \9 altogether after the fact) |
| 21:58 | <shu> | what in the world is \8 |
| 22:00 | <Bakkot> | an unfortunate way of spelling `8` |
| 22:03 | <shu> | what does a legacy octal escape do? |
| 22:04 | <Bakkot> | lets you write a code point by writing its code point value in octal |
| 22:04 | <shu> | does it produce the ascii code? |
| 22:05 | <shu> | well then |
| 22:08 | <rickbutton> | is there a standard way that people like to use to generate spec html in a PR for a proposal repo? |
| 22:09 | <ljharb> | rickbutton: it's basically just `ecmarkup path/to/emu/file` |
| 22:10 | <shu> | rkirsling: i wholeheartedly agree with michael ficarra's preference of making them errors |
| 22:10 | <ljharb> | rickbutton: if the proposal is generated from the template, the machinery is already there |
| 22:10 | <rickbutton> | right, i mean the hosting part, like the "put it on a webserver when a PR is opened" magic |
| 22:10 | <Bakkot> | I assumed the question was about getting previews |
| 22:10 | <shu> | rkirsling: but i also do not want to do the work |
| 22:10 | <ljharb> | rickbutton: set your repo's github pages settings to use the master branch |
| 22:10 | <ljharb> | rickbutton: and then `index.html` will Just Work |
| 22:10 | <devsnek> | rickbutton: in the olden days people used @alrra/travis-scripts but now i think people use gh actions |
| 22:10 | <shu> | rkirsling: the work of figuring out if anyone depends on this awful behavior |
| 22:11 | <ljharb> | rickbutton: see the steps here https://github.com/tc39/template-for-proposals#create-your-proposal-repo |
| 22:11 | <ljharb> | ahhh |
| 22:11 | <rkirsling> | shu: yeah that's the thing :-/ |
| 22:11 | <ljharb> | rickbutton: yeah if you want to *automatically* generate it you'd need something like an action |
| 22:11 | <ljharb> | rickbutton: devsnek: i've got one in progress for the template that auto-rebases PRs to include the updated spec text, fwiw |
| 22:12 | <rkirsling> | shu: I actually was only think of this in terms of surface tokens and wasn't dwelling on the fact that `"\7" !== "7"` so that makes me definitely feel like consistency with numeric literals isn't any sort of imperative here but |
| 22:12 | <rickbutton> | yeah I've set up something like netlify for this before, was interested to see if anyone else had done it, I can peek at other proposal repos, I feel like I've seen it before |
| 22:12 | <shu> | rkirsling: i don't see what consistency there is |
| 22:13 | <ljharb> | rickbutton: since github has github pages built in it seems unnecessary to add a dependency on an external service ¯\_(ツ)_/¯ |
| 22:13 | <rickbutton> | ljharb: you aren't wrong, especially with github actions |
| 22:14 | <shu> | rkirsling: i think i'm fine with forbidding \8 and \9 tbh |
| 22:15 | <shu> | rkirsling: we can query http archive for \8 and \9 |
| 22:15 | <rkirsling> | shu: yeah I think it was a naive miscategorization on my part. but if we can't collect data then I guess the only other option is "ship and see what fires arise" |
| 22:15 | <rkirsling> | ah that's fair |
| 22:15 | <shu> | well for this we *can* collect data right |
| 22:16 | <shu> | like it could come back there're no occurrences of \8 or \9, which seems likely to me |
| 22:17 | <rkirsling> | sure |
| 22:19 | <shu> | rkirsling: if you write some SQL and ping me to run it, sounds easy enough to figure out |
| 22:22 | <rkirsling> | shu: fair enough. I haven't actually looked at this before; now I get what the "don't run this" in that presentation was about 😂 |
| 22:24 | <rkirsling> | hmm, if it only gives an occurrence count, I wonder if we need to consciously filter out regexes? |
| 22:25 | <shu> | it doesn't have to give an occurrence count |
| 22:25 | <shu> | you can have it return the page URLs or the asset URLs |
| 22:25 | <shu> | and then look at them manually |
| 22:25 | <rkirsling> | ah cool |
| 22:25 | <shu> | the main problem is i don't know how to use GCP |
| 22:25 | <shu> | or any cloud platform, for that matter |
| 22:34 | <rkirsling> | yeah I'm looking at the tutorial |
| 22:34 | <rkirsling> | who did the stuff for the subclassing query? |
| 22:35 | <shu> | bradleymeck and myself |
| 22:54 | <rkirsling> | um so simply editing that query, I think this might be reasonable |
| 22:54 | <rkirsling> | https://www.irccloud.com/pastebin/LcAkatsE/ |
| 22:55 | <rkirsling> | I can try to run it myself and see if it yells about insufficient funds lol |
| 22:56 | <rkirsling> | oh whoa those two tables are like 20 TB together |
| 23:00 | <rkirsling> | shu: wdyt |
| 23:03 | <shu> | rkirsling: let's look at it tomorrow, no time today unfortunately |
| 23:03 | <rkirsling> | sgtm |
| 23:04 | <shu> | and yeah shit is huge |
| 23:06 | <shu> | i'm not *this* much of a capitalist, but suppose if we didn't have access to a corp-sponsored GCP instance and had to spend our own money to even begin asking a web compat question |
| 23:06 | <shu> | i wonder what our calculus will be |
| 23:07 | <shu> | "is changing behavior at least worth the $XXX in compute" |
| 23:14 | <ljharb> | so it'd be in the 3 digit range? |
| 23:15 | <ljharb> | (that's a genuine q, i actually have no idea how much it'd cost) |
| 23:22 | <rkirsling> | I'm seeing 5 USD / TB |
| 23:23 | <ljharb> | how many tibs does a web compat check typically take tho |
| 23:24 | <rkirsling> | seems like "20 if you don't make a mistake"? |
| 23:25 | <ljharb> | i mean, $100 to figure out if we can unmake a mistake in the ecosystem where mistakes can never be unmade seems like a bargain? |
| 23:26 | <ljharb> | like i feel like it wouldn't be hard to find a dozen developers somewhere who'd happily sponsor $100 to fix some little edge case |
| 23:26 | <rkirsling> | sure; to be clear, I meant a mistake in the query, since it's kind of like really expensive punchcards |
| 23:27 | <ljharb> | fair |
| 23:27 | <ljharb> | but i mean like, i feel like it'd be easy to raise thousands for something like this, which would give multiple tries |
| 23:27 | <ljharb> | i dunno tho, i've obv never tried it |
| 23:47 | <Bakkot> | personally I prefer spending Google's money if that's an option |
| 23:48 | <rkirsling> | thankfully that appears to be the present situation |