Pluto startup fails: "Uncaught TypeError: Cannot read properties of undefined (reading 'Library')"

[gjm-p]

I am using Windows 10, Julia 1.7.2, Pluto 0.19.1, and Chrome 1000.4896.127.

I have just installed Pluto by doing add Pluto at a Julia pkg prompt. When I then do Pluto.run(), a browser window opens displaying a Pluto welcome notice, saying "Loading...", and a progress bar at the top of that window starts growing. The progress bar eventually reaches the right-hand side of the window, but the "Loading" notice does not go away. If I click on, say, the "new notebook" link on the welcome page, I get a blank page but no content ever appears in it.

If I point a new browser window at the same URI as the original one launched by Pluto.run(), the same thing happens.

In the Javascript console, an error appears immediately on going to that URI:

Uncaught TypeError: Cannot read properties of undefined (reading 'Library')
    at n (editor.30fdc042.js:1:21495)
    at Object.<anonymous> (editor.30fdc042.js:1:21644)
    at r (editor.30fdc042.js:1:402)
    at Object.<anonymous> (editor.30fdc042.js:1:16652)
    at r (editor.30fdc042.js:1:402)
    at index.1a7d4001.js:1:677

I attempted to capture a video recording using the Windows "Game Bar" as requested, but even though it is only 27 seconds long and shows approximately nothing happening in a single window the file is 33MB in size. I do not have a YouTube account. I am fairly sure that the only further information it provides beyond what I have written above is about timing, so: After pasting the URI into the Chrome URL bar, it takes maybe 0.5s or so before (exactly simultaneously, so far as I can see) the welcome notice appears in the browser window and the "Pluto.jl, by ..." message appears in the Javascript console. After about another 0.1s, the error above appears, I-think-simultaneously with when the title in the browser tab changes from showing the URI to saying "Pluto.jl". Another 0.1s after that, the progress bar starts doing its thing; from this point on, nothing visible happens other than the movement of the progress bar. It moves rapidly (~3 seconds) to maybe 2/3 of the way across the window and then crawls more slowly (~10s) to the right-hand side. Nothing further changes after that.

Looking at where the exception happens in the devtools, it appears to be (after prettyprinting the minified JS) here:

let e = new window.observablehq.Library;

The thing that is squiggly-underlined is "Library" but I think what this actually means is that window.observablehq is undefined.

What I expected to see

Pluto welcome screen, actual ability to create or load notebooks, no errors in Javascript console.

What actually happened

Pluto welcome screen in "loading" state, immediate error in Javascript console, never emerges from "loading" state, cannot create or load notebooks.

[gjm-p]

In case it's useful, here's the output when I did add Pluto:

(@v1.7) pkg> add Pluto
    Updating registry at `C:\Users\GMcCaughan\.julia\registries\General.toml`
   Resolving package versions...
   Installed MsgPack ────────── v1.1.0
   Installed FuzzyCompletions ─ v0.5.0
   Installed Configurations ─── v0.17.3
   Installed ExproniconLite ─── v0.6.13
   Installed MIMEs ──────────── v0.1.3
   Installed Pluto ──────────── v0.19.1
    Updating `C:\Users\GMcCaughan\.julia\environments\v1.7\Project.toml`
  [c3e4b0f8] + Pluto v0.19.1
    Updating `C:\Users\GMcCaughan\.julia\environments\v1.7\Manifest.toml`
  [5218b696] + Configurations v0.17.3
  [55351af7] + ExproniconLite v0.6.13
  [fb4132e2] + FuzzyCompletions v0.5.0
  [6c6e2e6c] + MIMEs v0.1.3
  [99f44e22] + MsgPack v1.1.0
  [c3e4b0f8] + Pluto v0.19.1
Precompiling project...
  6 dependencies successfully precompiled in 9 seconds (137 already precompiled)

[gjm-p]

It looks to me as if (1) editor.30fdc042.js is the first Javascript to be pulled in, and (2) neither it nor the welcome page itself contains anything that would cause window.observablehq to be set. Is this some sort of Javascript bundling problem?

[fonsp]

Thanks for the report! It has now been fixed, see #2045

[gjm-p]

(I did check for pre-existing reports! But presumably 2045 was submitted between when I started writing mine and when I finished :-).)

I confirm that the problem does indeed appear to be fixed for me with 0.19.2. Thanks!