[adblocker-electron] Application is extremely lagged (memory leak)

Question

[adblocker-electron] Application is extremely lagged (memory leak)

JellyBrick opened this issue 9 months ago · comments

JellyBrick commented 9 months ago

7782d67
Performance regression occurred after that commit.

(This issue does not occur in versions 1.26.3 and earlier)

Code

https://github.com/organization/youtube-music-next/blob/feat/typescript/plugins/adblocker/blocker.ts

Related issues

th-ch/youtube-music#1187
th-ch/youtube-music#1208
th-ch/youtube-music#1212

Rémi · Answer 1 · Sun Sep 17 2023 00:05:39 GMT+0800 (China Standard Time)

Hi @JellyBrick,

Thanks for reaching out and sorry for the delay in answering. This looks like a serious issue and multiple people have already encountered it. At the moment I have limited bandwidth (and limited knowledge about Electron stack) so I am not able to investigate it myself. I'm gonna ping a few people who are familiar with it in the hope that we can get some clarity regarding the root cause of the slow-down.

@DrRoot-github since you opened the PR and are most likely the person with most insights in the change, would you have some availability to look into it? I believe @Jelmerro also investigated as part of the issue on Vieb-side.

Could this issue be caused by using sendSync from the page? Alternatively, maybe we are returning too much on updates (e.g. maybe we wrongly return scriplets or too many selectors).

I hope that helps,

DrRoot · Answer 2 · Sun Sep 17 2023 02:51:05 GMT+0800 (China Standard Time)

I'll try to investigate a bit more, so please wait a little longer. If possible, it would be helpful if you could provide specific instructions on how to reproduce it. I think I might be the culprit, seriously sorry.

Rémi · Answer 3 · Sun Sep 17 2023 19:21:43 GMT+0800 (China Standard Time)

@DrRoot-github No worries at all, I think I found the issue and created a PR to fix it here: #3441 Would you have time to review it and maybe test that the blocking of ads in YouTube didn't regress? (from some local testing it seems fine).

Cc @Jelmerro once reviewed we can publish an update of the adblocker package (as early as today if anyone has time to review the changeset).

KunaPrime · Answer 4 · Sun Sep 17 2023 22:58:52 GMT+0800 (China Standard Time)

I'm user of Vieb and @Jelmerro have provided me a patch with a fix (i assume it is based on #3441) and i can confirm that regression is gone.

Rémi · Answer 5 · Sun Sep 17 2023 23:05:48 GMT+0800 (China Standard Time)

Thank you @KunaPrime and @Jelmerro, I have merged the PR and a new release will be available shortly.

Rémi · Answer 6 · Mon Sep 18 2023 00:30:05 GMT+0800 (China Standard Time)

The release pipeline got broken by some upstream dependency updates. Until it gets fixed, I have pushed a canary release manually which you can try out: https://www.npmjs.com/package/@cliqz/adblocker-electron/v/1.26.7--canary.50688cf.0

JellyBrick · Answer 7 · Tue Oct 17 2023 20:46:14 GMT+0800 (China Standard Time)

This issue is still reproducible in v1.26.8

Rémi · Answer 8 · Wed Oct 18 2023 02:55:43 GMT+0800 (China Standard Time)

Could you provide some more details about this please?

JellyBrick · Answer 9 · Wed Oct 18 2023 17:16:40 GMT+0800 (China Standard Time)

I found the problem that the responsiveness of the webpage has decreased over time.

(Electron 27.0.0)

Tested webpage: https://music.youtube.com