The behavior is weird when changing the "Save Location"

[jakobjw]

Environment: macOS
Userscripts Safari Version 4.8.3 (131)

In the "Editor Settings", after selecting a new "Save Location" (via the "gears" button and the macOS folder selector), the following happens:

• The displayed path for "Save Location" is still the old one (doesn't get updated)
• Clicking once more on the "gears" button however shows the folder for the updated (new) path

Only after re-opening the Extension Page, the new updated "Save Location" is correctly displayed.

My old path was a local path and the new path an iCloud Drive path.

---

As a side note, tbh I also would have expected a prompt if the existing userscript files shall be moved/copied to the new location.

[ACTCD]

The related experience has been improved in the v5.0.0 beta version.

You can try it out via TestFlight:
https://github.com/quoid/userscripts/releases

As a side note, tbh I also would have expected a prompt if the existing userscript files shall be moved/copied to the new location.

So we never copy or move your files.

[jakobjw]

Ok thanks for the hint! Just tested the v5.0.0-beta.20 and there this behavior is "not weird". 👍

So we never copy or move your files.

Makes sense.

---

Ok cool, can be closed from my side of view! 🚀

[ACTCD]

Actually, double checking again, this improvement should have made it into v4.x.x:

• Improve the interaction between app and extension #404

That said, the expected behavior is to auto refresh the extension page after actually changing the path.

It actually still works, but only with a small chances, and it seems like most of the time it doesn't work.

But in beta it always works fine.

I need to do more check and debug to see what is causing this probabilistic error.

Thanks for the feedback and raising it, otherwise I wouldn't have noticed the issue in v4.x.x.

[ACTCD]

After countless attempts and difficult investigations.

I can almost confirm this is a bug in upstream WebKit engine, and it's related to code minification.

I don't know if it's a listener/event related implementation bug, or a GC related bug.

The issue appears to be that the listener is not called, but interestingly it may become valid or invalid due to subsequent registration of new listeners.

This was the tricky part that led me to spending a lot of time and still not being able to find a way to reproduce it reliably. But only able to increased the probability of reproducing the error.

Minimized code that easily triggers this bug:

function f1() {
	return browser.runtime.connectNative();
}
f1().onMessage.addListener((message) => {
	console.info(1, message);
});

Anyway, in the beta version, the minimizer generated different code, causing the bug to not seem to be triggered:

function f2() {
	return browser.runtime.connectNative();
}
const p = f2();
p.onMessage.addListener((message) => {
	console.info(2, message);
});

You can see there's an intermediate variable there.

---

And this issue actually related to another unresolved listener related bug I observed maybe 1-2 year ago.

• For example, turning off the switch in the popup will not synchronize the status to the already opened extension page. But it was designed to be reactive, and it did works fine sometimes.

The issue also doesn't seem to occur in beta builds, but the minimized code is not the difference mentioned above. I did not continue to investigate the discrepancies that triggered the issue.

---

Anyway, there's not much I can do there.

And since both issues don't seem to exist in the beta, maybe newer minification and build tools fixed these bugs.

I'm reluctant to update the build environment and dependencies for v4.x.x, that's a lot of work. And not worth it.

Cause I need to prioritize my time investing in the refactoring of v5.0.0.

And these issues will be resolved once v5.0.0 is officially released.

I will close this issue as won't fix.

[jakobjw]

I can almost confirm this is a bug in upstream WebKit engine, and it's related to code minification.

Wow, crazy! 🤯😅
Thanks for spending so much effort looking into this! I didn't assume to cause this amount of work. 🙈

Just out of curiosity: Is the code minification stuff part of the closed-source Safari code? Or is it open-source somewhere?