Bug 255402 – Web Inspector doesn't show nested workers

• Component: Web Inspector
• Status: RESOLVED FIXED

[Thomas Steiner]

Created attachment 465890 [details]
Safari Console

A Web Worker can spawn other Web Workers; these are sometimes referred to as subworkers or sub-workers. Safari does not seem to request the subworker file and only spawns a Worker from the main thread.

Note the screenshots:

chrome.png: The expected behavior in Chrome.
firefox.png: The expected behavior in Firefox.
safari-console.png: Safari only logs the message from worker.js.
safari-network.png: Safari does not seem to make the request for subworker.js.
safari-sources.png: Safari does not show subworker.js in its sources.

Minimal repro: https://safari-subworker.glitch.me/

[Thomas Steiner]

Created attachment 465891 [details]
Safari Network

[Thomas Steiner]

Created attachment 465892 [details]
Safari Sources

[Thomas Steiner]

Created attachment 465893 [details]
Chrome

[Thomas Steiner]

Created attachment 465894 [details]
Firefox

[Thomas Steiner]

This does _not_ work in:

- Safari Tech Preview Release 167 (Safari 16.4, WebKit 18616.1.8.2)
- Safari Version 16.5 (18615.2.5.11.2)

(I'm on Ventura 13.4 Beta (22F5037d))

[Alexey Proskuryakov]

Nested workers were implemented in bug 22723, quite surprising that this isn't working.

[Karl Dubost]

And this is passing. 
https://wpt.fyi/results/workers/semantics/multiple-workers/003.html?label=experimental&label=master&aligned


And the example given in the spec.
https://html.spec.whatwg.org/multipage/workers.html#delegation
is working
https://html.spec.whatwg.org/demos/workers/multicore/page.html

Maybe it's about something else. Is it about the console.log()?

[Karl Dubost]

I even wonder if it's in fact just a Web Inspector issue.
a bit like Bug 252685

[Dasa Paddock]

That sounds right. I see that https://safari-subworker.glitch.me/subworker.js does get loaded when I test https://safari-subworker.glitch.me and have Charles Proxy running.

[Karl Dubost]

I renamed the bug.

[Sam Sneddon [:gsnedders]]

On Safari 16.4, macOS Ventura 13.3.1 (22E261), MacBook Pro 14" 2023 (Mac14,9), I am seeing the "from subworker" console messages, though I can't see anything in the Sources or Network tabs.

Given I can see them on relatively fast hardware, this makes me wonder if there's some race happening somewhere?

[Thomas Steiner]

I have added a bunch of postMessage() calls to the demo at https://safari-subworker.glitch.me/, so we can see communication happening in all directions. I can indeed confirm that the worker.js postMessage() call arrives in subworker.js, since its response is logged in worker.js.

## Here're the logs from Chrome:

Hello from main thread
worker.js:1 Hello from worker.js
worker.js:4 Message received in worker.js: From main thread to worker.js
script.js:6 Message received in main thread: From worker.js to main thread after message from main thread
subworker.js:1 Hello from subworker.js
subworker.js:4 Message received in subworker.js: From worker.js to subworker.js
worker.js:11 Message received in worker.js: From subworker to worker.js
script.js:6 Message received in main thread: From worker.js to main thread after message from subworker.js


## Here're the logs from Firefox:

Hello from main thread script.js:1:9
Hello from worker.js worker.js:1:9
Message received in worker.js: From main thread to worker.js worker.js:4:11
Message received in main thread: From worker.js to main thread after message from main thread script.js:6:11
Hello from subworker.js subworker.js:1:9
Message received in subworker.js: From worker.js to subworker.js subworker.js:4:11
Message received in worker.js: From subworker to worker.js worker.js:11:11
Message received in main thread: From worker.js to main thread after message from subworker.js script.js:6:11

​
## Here're the logs from Safari (Safari Tech Preview and Stable agree): 

[Log] Hello from main thread (script.js, line 1)
[Log] Hello from worker.js (worker.js, line 1)
[Log] Message received in worker.js: – "From main thread to worker.js" (worker.js, line 4)
[Log] Message received in main thread: – "From worker.js to main thread after message from main thread" (script.js, line 6)
[Log] Message received in worker.js: – "From subworker to worker.js" (worker.js, line 11)
[Log] Message received in main thread: – "From worker.js to main thread after message from subworker.js" (script.js, line 6)


This means it's indeed a complete Web Inspector blindness to anything happening in subworker.js. 

Also, it doesn't yet explain why SQLite Wasm backed by the OPFS (https://sqlite-wasm-opfs.glitch.me/) isn't working on Safari yet, but that's a different story then (maybe). I'll keep investigating. The subworkers theory was a good lead, though…

[Thomas Steiner]

Found the root cause of the SQLite Wasm problem: `navigator.storage.getDirectory()` fails in subworker with an `InvalidStateError`: https://bugs.webkit.org/show_bug.cgi?id=255458.

[Radar WebKit Bug Importer]

<rdar://problem/108322385>

[Devin Rousso]

Pull request: https://github.com/WebKit/WebKit/pull/13614

[Alexey Proskuryakov]

*** Bug 255637 has been marked as a duplicate of this bug. ***

[Alexey Proskuryakov]

*** Bug 272382 has been marked as a duplicate of this bug. ***

[EWS]

Committed 279793@main (b5d3f9b6724b): <https://commits.webkit.org/279793@main>

Reviewed commits have been landed. Closing PR #13614 and removing active labels.