Argh.. GitHub is broken again

[jrfnl]

Yup... GH has done it again.... must be driving you bonkers..

I see this error in the console:

TypeError: (intermediate value).attachShadow is not a function[Learn More] chunk-index2-a8e52977.js:1:2276

As for symptoms:

• Drop down on a comment (to edit the comment, for instance) not working.
• No notification showing for a newly pushed branch on the "Pull" page.
• etc

[Polynomial-C]

Same with the user drop down menu in seamonkey. All I can see in the menu is: Sorry, something went wrong.

[JustOff]

Resolved by 1.2.12.

But now GH is slow like a slug :angry:

[dilworks]

Almost there!

First, thanks for the speedy fix, GiggityHub must be driving you, me, and everybody else not using Holy Chrome®™ downright insane with this pointless chase for shiny.

Now the editor seems to work properly again, and dropdown menus work... except for one: the user menu. The "something happened" message is gone, but all I get is a blank menu instead.

EDIT (yay, I can edit again!): no errors are logged in console when trying to access the broken user menu.

[JustOff]

@dilworks, for me it works just fine:

[dilworks]

Weird... Let me test on a clean profile, clearing cache, etc.

[dirkf]

GH doesn't seem to be serving the code that provoked the error with missing this.attachShadow now, but kudos for instant 1.2.12.

[jrfnl]

@JustOff Your work is soooooo much appreciated. Thanks (again) for the quick fixes and releases ❤️

[JustOff]

@dirkf, perhaps there is some A/B test, but I still can see the error with 1.2.11.

[dirkf]

No, you're right. DevTools crashed the browser and I forgot to re-enable JS.

[AroKol78]

such a trifle - ~string~ text position (no bugs in the console) https://github.com/JustOff/github-wc-polyfill/commit/727fe0e5da881ceac90248f03b20a10b10e9f719

---

moving the cursor in the text with the keys - very slow

[dilworks]

@JustOff : Managed to reply the user menu issue on a totally clean SeaMonkey profile - no addons except for github-wc-polyfill, no custom settings, no nada:

All I can see on the network inspector is that the menu is loaded via AJAX (/menu) on the initial page load, but it's not being used for whatever reason.

[JustOff]

@dilworks, well, I can reproduce it with SeaMonkey, but not with Pale Moon. I wish I know what to do with this.

@AroKol78, yep, it's a bummer.

[dilworks]

@JustOff FWIW it also happens on SeaMonkey latest beta, 2.53.11b1...

[Vangelis66]

But now GH is slow like a slug 😠

I can confirm/reproduce in latest St52 32-bit 😠 ; typing this very comment is sluggish as hell 💢 , just reloading a GitHub tab now produces the Not Responding hang in the Browser Window title for 2-5 sec (YMMV), on "heavy" GitHub pages I may even get this:

So, M$ have now rendered GitHub very "uncomfortable" to use on "legacy" platforms 😭 ; after all, today is Feb 17th, I guess someone there did stick to their original plan to mess things up for "us" on the 17th...

[dirkf]

The top-right pull-down is an <include-fragment> element, which is where I was seeing the this.attachShadow is not a function error without 1.2.12 (and even though the polyfill from #19 was loaded and this.attachShadow was a function when the problem line was BPed and the page reloaded).

Anyhow, there's a big difference between adding APIs and redefining the language syntax (ES2020 optional chaining, ES2018 regex literals).

[JustOff]

even though the polyfill from https://github.com/JustOff/github-wc-polyfill/issues/19 was loaded and this.attachShadow was a function

This is because they need Element.prototype.attachShadow, not only HTMLElement.prototype.attachShadow

[Vangelis66]

https://github.com/JustOff/github-wc-polyfill/commit/727fe0e5da881ceac90248f03b20a10b10e9f719

(no bugs in the console)

I get

Content Security Policy: The page’s settings blocked the loading of a resource at self (“script-src 'sha256-adiMt0jngKHacvEid386+kd4GfsJG9Bk63VtPqP2ZVI=' https://github.githubassets.com”). Source: parent["VM0.12818166828098088"] = [this].

immediately after https://github.githubassets.com/assets/diffs-4fa71244.js is loaded...

[dirkf]

... they need Element.prototype.attachShadow ...

In fact I had modified the polyfill to target Element: perhaps I needed to do both given the rickety OO system in JS.

Also, what on earth are they doing in this typing field? Looks like typing up stuff in the editor and pasting it in will be the new MO.

[JustOff]

@Vangelis66, can't reproduce in a clean profile, perhaps you have some userscript ot other extension enabled for GH.

[Vangelis66]

can't reproduce in a clean profile, perhaps you have some userscript or other extension enabled for GH.

Apologies 😊 , I had indeed 3 GH-targeting userscripts enabled in Violentmonkey...

BTW, with 1.2.12 already installed, "Load more..." (items) is also broken 😢 :

[JustOff]

"Load more..." (items) is also broken

Can't confirm, works just fine in Pale Moon 29.4.4.

[AroKol78]

https://github.com/JustOff/github-wc-polyfill/issues/44#issuecomment-1043223634 i don't have such csp - if you press "load diff" it loads fine (just a strange text position) https://github.com/JustOff/github-wc-polyfill/issues/44#issuecomment-1043251878 is fine, only there is a problem with reloading

[dirkf]

'Load more' working OK, if slow, in SM 2.53.10.2 with 1.2.12 installed.

[Vangelis66]

Can't confirm, works just fine in Pale Moon 29.4.4.

😊 ; I somehow made it work, too (and now hid my invalid comments); I probably need to go out and get some fresh air; apologies to all, sorry 😜

[garoto]

Wow, JustOff weren't exaggerating when he mentioned its slowness, damn. GH is pretty much unusable now on latest PM.

I had to type this comment in my text editor, since the comment form was lagging like crazy with one core at 100%. Nuts!

[dilworks]

Huh, I always thought the GH editor was always slow (and sometimes buggy), but today indeed has become VERY laggy for me, like something that is doing deep learning on every keystroke...

The modern web is a dumpster fire of broken tech :/

[AroKol78]

script not responding https://github.com/ikatyang/emoji-cheat-sheet

[SeaHOH]

I try to replace chunk-index2 to last work version a3fdc9f7, it works. But may some of site functions will be broken.

Patch in my fork (not for SM): https://github.com/SeaHOH/github-wc-polyfill/commit/f5c231650285a4f3c8359bb49b9d59313e94b703

The new additional code of chunk-index2 is:

constructor(){super();this.attachShadow({mode:"open"}).innerHTML=`
      <style> 
        :host {
          display: block;
        }
      </style>
      <slot></slot>`}

[AroKol78]

I'm back to 1.2.11 (I chose the lesser evil)

[ghost]

Resolved by 1.2.12.

But now GH is slow like a slug 😠

I'm back to 1.2.11 (I chose the lesser evil)

So, what to do? In actual state 1.2.12 fully blows CPU and on my laptop it always just stuck — Pale Moon asks me to stop script instead...

EDITED: Just updated to 1.2.13b2 (https://github.com/JustOff/github-wc-polyfill/issues/44#issuecomment-1044231249) — seems like all is work as for now.

Thanks, @JustOff!

[Vangelis66]

Thanks for this kludge 😄 , SeaHOH 👍 ; installed it in latest Serpent 52 and GH is again usable (typing this from St52 right now...); BTW, no disrespect is meant to the great JustOff 👍 ; this fix will probably get us going till tomorrow, when M$, no doubt, introduce more Chrome-specific "goodies" 😠 ...

[JustOff]

Patch in my fork (not for SM): https://github.com/SeaHOH/github-wc-polyfill/commit/f5c231650285a4f3c8359bb49b9d59313e94b703

Unfortunately, it doesn't work for me. Perhaps I'm missing something.

[dirkf]

Is there a high-numbered TV channel somewhere showing conspiracy docs about how M$ and Google are going to take over the world and split it up between them? I really don't see why M$ would want to promote content that favours G. I favour the view that M$ is generally hands-off with GH but the GH staff making detailed technical decisions love the shiny to the detriment of configuration control and without a mature understanding of their implications.

More significantly, with 1.2.12, I can type again but the 'TypeError: this.attachShadow is not a function' diagnostic is back, the comment top-right ... button gives an empty pull-down and the page top-right user pull-down says 'Something went wrong'. (How true) Dropped back to 1.2.10: ludicrously slow typing but comment menus back. And now the original behaviour with 1.2.12 is back, with empty page top-right user pull-down.

[SeaHOH]

Patch in my fork (not for SM): SeaHOH@f5c2316

Unfortunately, it doesn't work for me. Perhaps I'm missing something.

You patch your version 1.2.11? How about version 1.2.10? Replace environment with chunk-index2.

[dirkf]

Possibly this is the reason for the slowdown:

queueMicrotask.toSource()
/*
(function(f) {setTimeout(f, 0)})
*/

I don't know whether that's being sent by GH or set by 1.2.12, but it should be polyfilled with s/t like this:

const _queueMicrotask = (function () {
    const checkIsCallable = (callback) => {
        if (typeof callback !== 'function') {
            throw new TypeError( "Failed to execute 'queueMicrotask': callback is not a function" );
        }  
    };
    let promise = null;
    return function(callback) {
        checkIsCallable(callback);
        (promise || (promise = Promise.resolve()))
            .then( () => callback() ) // call with no arguments
            // if an error occurs during callback execution,
            // throw it back to globalObj (using setTimeout to get out of Promise chain)
            .catch( (err) => setTimeout( () => { throw err; } ) );
    };
})();

for (let intf of [window, Worker.prototype]) {
    if (!('queueMicrotask' in intf)) {
        intf.queueMicrotask = _queueMicrotask;
    }
}

[SeaHOH]

Yep, I only tested my fork before. Now, I tested that only replace the chunk-index2 is not enough. And I really use Promise to polyfill with queueMicrotask in my fork, and kept the attribute defer when replace script elements.

[JustOff]

kept the attribute defer when replace script elements

Yep, that was the part I missed, and here it is: 1.2.13b2.

[JustOff]

FTR, SeaMonkey 2.53.9.1 has no problems with the user menu with both 1.2.12 and 1.2.13b2, while 2.53.10 has.

[SeaHOH]

I forgot data-module-id, maybe the reason is it.

[dirkf]

Certainly I don't see a relevant change in the SM 2.53.10 changelog: it's all ChatZilla, MailNews and browser chrome stuff.

[RamonUnch]

Thanks a lot @JustOff and @SeaHOH. Thanks to you I do not have to use chrome for GitHub. I deeply appreciate your work. The 1.2.13b2 works very well and fast for me.

[Polynomial-C]

User drop down menu still is empty with 1.2.13b2 and seamonkey-2.53.11b1.

[JustOff]

You could try to find a regression between 2.53.9.1 and 2.53.10 based on these nightly builds.

[garoto]

Noticing some subjectively faster GH performance overall with 1.2.13, specially when rendering somewhat larger diffs and issues with 100+ comments. e: with latest PM release on Windows.

[hiiamboris]

Is there a high-numbered TV channel somewhere showing conspiracy docs about how M$ and Google are going to take over the world and split it up between them? I really don't see why M$ would want to promote content that favours G.

@dirkf Just look at the top stakeholders list of Google/Alphabet, M$ or really any other Silicon Valley corp, you'll be in for a surprise ;)

[garoto]

When a corporation reaches the multi-billion dollar mark in valuation and/or assets, it can't call its shots anymore. It either play ball or bad things will happen. Specially true for the computing industry for the past two decades or so.

[CyberTailor]

@JustOff

Resolved by 1.2.12.

But now GH is slow like a slug

Let's move this repository to some Gitea public instance (like https://codeberg.org) so we can report issues at least. It works perfectly in Pale Moon.

[rofl0r]

Let's move this repository to some Gitea public instance (like https://codeberg.org/) so we can report issues at least. It works perfectly in Pale Moon.

good idea, if you s/move/copy/

[CyberTailor]

good idea, if you s/move/copy/

GH can be left as a mirror, indeed.

[JustOff]

Let's move this repository to some Gitea public instance good idea, if you s/move/copy/

Sorry, but this is unlikely to happen.

so we can report issues at least

You can always use a different browser for this.