✨ Launch minimal-wrapper native CEv1 #26360
Conversation
|
Delta in bundle size expected? |
|
Yes, it's expected. We're now including the minimal polyfill and the checks necessary to find which polyfill to install. |
|
Ping @cramforce and @choumx |
Sigh, I can't rebase on upstream.
| @@ -958,7 +958,7 @@ export function install(win, ctor) { | |||
| const {Reflect} = win; | |||
|
|
|||
| // "Construct" ctor using ES5 idioms | |||
| const instance = Object.create(ctor.prototype); | |||
| const instance = /** @type {!Function} */ (Object.create(ctor.prototype)); | |||
There was a problem hiding this comment.
I'm not really sure, but closure doesn't like the Function.call.call() below without it.
| */ | ||
| export function install(win, opt_ctor) { | ||
| export function install(win, ctor) { |
There was a problem hiding this comment.
Why not just remove ctor parameter?
There was a problem hiding this comment.
Because we still need to detect if we're in ESM (real classes) or ES5 (transpiled classes).
There was a problem hiding this comment.
Can be inlined in install() which would make the API a bit cleaner (optional).
| win, | ||
| NATIVE_CUSTOM_ELEMENTS_V1 ? class {} : undefined | ||
| ); | ||
| installCustomElements(win, class {}); |
There was a problem hiding this comment.
We always transpile our testing code right? Did we try running our test suite against the native CE code path?
There was a problem hiding this comment.
I'm not sure if the ESM builds ran with this RTV experiment enabled. @erwinmombay?
|
We observed no latency impact at p95 and p50 for first contentful paint (or any metric), and no new unusual errors. |
|
Was there any impact for First-Viewport-Ready?
…On Wed, Jan 22, 2020 at 2:26 PM Justin Ridgewell ***@***.***> wrote:
We observed no latency impact at p95 and p50 for first contentful paint,
and no new unusual errors.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#26360?email_source=notifications&email_token=AAAV4T3AJPQUTI5PTKRD5KTQ7DBY5A5CNFSM4KHKTBE2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEJVKRYI#issuecomment-577415393>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/AAAV4T346VABPEOZNTDUWMLQ7DBY5ANCNFSM4KHKTBEQ>
.
|
|
No, everything was completely stable. |
* master: (62 commits) 📦 Update dependency fetch-mock to v8.3.2 (ampproject#26491) Revert 'Move mutator implementations out to a standalone service' (ampproject#26479) Fix syntax error (ampproject#26481) Add pespective back. (ampproject#26486) More user friendly errors in layout.js (ampproject#26448) ✨ Start logging AMP URL on SwG Pages (ampproject#26480) Fix border around desktop amp-story-pages. (ampproject#26449) Fix Story tests. (ampproject#26464) ✨ Performance Measurement Chrome Extension (ampproject#26333) amp-consent restrict iframe fullScreen if no focus (ampproject#26461) Add performance benchmark task (ampproject#26026) ♻️ amp-script: emit warning if zero height and width. (ampproject#26444) ✨ Launch minimal-wrapper native CEv1 (ampproject#26360) ♻️ Lint: include externs (round 2) (ampproject#26446) amp-script: Create 'fill content' container for responsive/fluid (ampproject#26400) amp-consent remove cmp iframe focus (ampproject#26437) Disable macro-after-long-task in inabox. (ampproject#26459) Launch layoutbox-invalidate-on-scroll (ampproject#26430) Add amp-ad support for ByPlay (ampproject#25663) 🏗 Add specific RTV opt-in to experiments.html (ampproject#26434) ...
This pushes us closer to use real, native CEv1!
Now, instead of forcing a full polyfill in every browser, we will have the following support matrix:
falsefalsefalsetruetruefalseHTMLElementtruetrueNote that we won't be able to eliminate the polyfill from module builds, since Firefox supported modules before CEv1. 😢