packages: adblocker-electron{-preload}: make use of executeJavaScript

[heaptr]

Use WebContents.executeJavaScript for adblocker script injection.

Preload script injection is subject to CSP, breaking functionality on some websites.
This change uses executeJavaScript API to bypass CSP:

• Replace preload injection with IPC to main process
• Add adblocker-inject-script handler in main process
• Throw an error in the renderer process if the injected script fails

Fixes #4234

[chrmod]

can we inject scriplets directly from get-cosmetic-filters-first handler?

this will avoid unnecessary communication as what is happening here is:

• 1. preload requests scriptlets by get-cosmetic-filters-first
• 2. background matches the scritlets and sends it back to preload
• 3. preload sends scriptlets back to background to perform the injection with adblocker-inject-script
• 4. background injects scripts in the main world

We can avoid steps 2 and 3, but injecting the scriptlets directly in the step 1.

[heaptr]

@seia-soto @chrmod I tried to streamline the implementation and fix the issues, please take another look at the patch.

[seia-soto]

Thank you for the contribution. I found some headrooms to improve, but I really appreciate introducing the changes.

[heaptr]

@seia-soto Thank you for the review! PTAL. I'll squash the commits once the review is through.

[seia-soto]

Overall looks very great. No other comments from me. I requested a review from our team.

[seia-soto]

I found that we were missing tests from adblocker library from #4302 but this PR doesn't have any relationship with that. Therefore, I think it's fine to continue without additional rebasing.

[chrmod]

Thank you very much for pushing this forward. It is clear that electron deserves improvements.

This review may take a bit more time as the entire team is catching up with electron changes and we want to be sure we can get most out of your efforts, so please bear with us :)

Was thinking, how about instead of requesting scriptlets from the preload script, we push them directly based on did-navigate and did-frame-navigate which would make it closer to the way the Ghostery extension does the scritplet injection

With that approach, changes to the other cosmetic filters injection could be minimal.

[heaptr]

Thank you for the review @chrmod! I've replied most of the comments, and I think the most important one to address is that DOMUpdater expects a sync function.

I’d prefer to handle broader changes in other PRs since the electron package isn't fully functional right now. We rely on @ghostery/adblocker-electron in our product, so we'll definitely be here to help maintain the project as issues arise.

[heaptr]

@chrmod PTAL, updated the branch!

[chrmod]

We're almost there :)

Still need to ensure everything happens in the right moment.

Please let me know if you have any questions.

[heaptr]

@chrmod Pushed some changes that hopefully addresses the review. I'm losing the thread a bit, please let me know if the changes conform to what you suggested.

[chrmod]

I think we need one more change around the async code.

Can you already confirm that this works with youtube.com ads?

[heaptr]

@chrmod Committed the suggestions. I can confirm that this works on youtube.com.

[chrmod]

Looks like one last thing is missing - I've made a code suggestion

[chrmod]

This is now merge ready, we only have to fix the linter.

[seia-soto]

Running CI. 👍

[seia-soto]

We would like to extend our sincere appreciation and congratulations to you for your valuable contribution to the adblocker library. We are pleased to merge this fix, which will benefit all users, and we are grateful for the time and effort you invested in addressing this issue.

[chrmod]

Many thanks @nullptropy !

[heaptr]

Thanks for being patient with me and for all the hard work you guys put into this library, I really appreciate it! @seia-soto @chrmod

[PascalPixel]

wonderful work, thank you for your time and effort (ﾉ´ヮ´)ﾉ*: ･ﾟ