fix broken copy, paste and showFindMode since Chrome 74

[gdh1995]

During tests I found that a setTimeout is needed, although I haven't found why.

This can fix #3260, fix #3292, fix #3297, fix #3298, fix #3302, fix #3307, fix #3311, fix #3313 and fix #3315 .

[gdh1995]

@smblott-github Would you like to merge this for just a try? There're quite a few issues reporting the new FindMode and clipboard issue on Chrome 74, and this PR works on my laptop and PC (both are Win 10 x64).

Update: I've confirmed the detailed reason of this breaking and this PR should work well on a very slow computer.

[riftEmber]

How did you divine the 17ms solution? (I haven't tested)

[gdh1995]

How did you divine the 17ms solution? (I haven't tested)

I once thought it's the recently-added "user activation" logic (since Chrome 72) that prevents Vimium code from working. The "user activation" logic is designed to pass privileges across frames and remove them after a while. Therefore, if it has bugs, Chrome may mistakenly pass "disabled" status from a normal web page to Vimium's privileged frames, and then at a small time span code may fail.

So I tried adding a timeout and found the privilege came back. Recently, I noticed that "0ms" might be not enough, and "17ms" seems to work well - if not, we may increase it again.

But I navigated Chromium source code for minutes but didn't find any buggy source code. More survey is needed.

[gdh1995]

Update: after more tests, I find it's a race condition:

• the action of element.focus() / select() can work only if a document "is" visible
  • if a doc is not visible, getSelection().selectAllChildren() works but
  • the "visible selection" is none, and getSelection() + "" always returns "" (an empty string).
• But in most time a HUD iframe has vimiumUIComponentHidden and its "display" is "none",
  • although Vimium code may call "toggleIframeElementClasses" for paste
  • it needs a tiny time span to spread the "display:block" style to the HUD frame.

Therefore, a toggleIframeElementClasses + getComputedStyle before postMessage can help a lot:

• should call toggleIframeElementClasses before all 3 HUD actions: showFind, copy, paste
• getComputedStyle is to enforce a style update immediately

But during tests in today there're about 1 / 10 cases where th HUD could not get focused, so I think the setTimeout 17 is still necessary.

[innaterebel]

Let's also try tagging @philc to see if he will sort things out. (if tagging the project owner is even necessary for him to get notified...)

[gdh1995]

I've confirmed the comment I wrote above:

• Chrome 74 refuses Element.focus and Selection.toString in an iframe if the iframe is not visible
• Vimium's iframePort for HUD iframe is created by MessageChannel
  • it sends messages immediately when postMessage is called, and stack trace is:
  • https://cs.chromium.org/chromium/src/third_party/blink/renderer/core/messaging/message_port.cc?dr=C&g=0&l=76
  • https://cs.chromium.org/chromium/src/mojo/public/cpp/bindings/lib/connector.cc?dr=C&g=0&l=315
• changes to <iframe>.classList posts a new task and Chrome does style updates in the task
  • the task is not executed until the JavaScript code finishes and the task manager continues its task loop.
  • during the task, some messages are sent from parent frame (a content page) to child frame (HUD)
  • the message channel is IPC
  • one message is RenderWidget::OnWasShown and stack trace on HUD is:
    • RenderWidget::OnWasShown -> RenderWidget::SetHidden -> LayerTreeView::SetVisible
• then there's a potential race condition between execCommand("Copy") and SetVisible
  • in fact, usually execCommand("Copy") is eariler if only HUD has been inited
  • while before Chrome 74 a hidden frame of display:none can still copy text
    • making us ignore the race condition.

Here's the result of chrome://tracing using Vimium master (v1.64.5) on Chrome 74.0.3729.108 x64 (blank User Data, Win 10):

The result of this PR:

[mohitt]

Great tool :). But this bug, after update is causing huge pain. Please update the extension

[gdh1995]

@smblott-github Sorry to disturb you again, but there have been so many issues reporting FindMode focusing and copying problem on Chrome 74. Would you have time to merge this? This PR has been verified carefully.

[funnydman]

Guys, please merge it if everything is okay because someone switched to the mouse again(

[athakral]

If someone is in dublin, Ireland. May be he can request him personally.

https://www.computing.dcu.ie/~sblott/

[megalithic]

@smblott-github friend; this is huge! would love to have this merged in, to further make this tool amazing again! thanks!

[elongl]

Why don't we merge this is it seems to work?
Also, isn't there a better solution than setting a timeout?

[gdh1995]

@elongl The timeout is just in case that a computer might be too slow, and won't generate appreciable delay, so I think it should be here (The delay of loading Vomnibar is mostly from common HTML webpage loading and painting).

[elongl]

So why don't we merge?

…

On Wed, May 8, 2019 at 3:41 PM Dahan Gong ***@***.***> wrote: @elongl <https://github.com/elongl> The timeout is just in case that a computer might be too slow, and won't generate appreciable delay, so I think it should be here (The delay of loading Vomnibar is mostly from common HTML webpage loading and painting). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#3277 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AHGIP4PN6Z2N7IOLYBHGSNLPULC6NANCNFSM4HAUFNYQ> .

[adriangonzalez]

For anyone that's getting a little jumpy on this issue, it looks like a fix is coming, but in the meantime, cmd/ctrl+F still works ;)

[innaterebel]

Hate to break it to you guys, but do you know the word "abandoned"? That's where we indeterminately are.

While everybody is welcomed to make pull requests in an open source project, a project does have an owner and owner-approved committers, who are the only people possessing the privilege to actually merge a pull request and publish an official build.
For Vimium, the owner is @philc, and currently the only committer is @smblott-github. Unfortunately, both of them have been inactive for quite some time for reasons unknown, so the practically abandoned state we are in.

The abandonment is likely not intentional, but there is no guarantee the two will come back. If you are really bothered by the issues and cannot wait around it, you may want to find an alternative extension.

[gdh1995]

@mrmr1993 Sorry to disturb you but do you still have committing permission to Vimium repo? I'd like this PR to be merged to fix Vimium's FindMode focus and clipboard operations issues, which occur on Chrome 74 and are quite annoying.

[mrmr1993]

@gdh1995 never had commit permissions. I left a query comment on the code, but nothing I can do on actually merging, sorry.

[ssangervasi]

innaterebel was a bit sassy, but it is likely the maintainers have lost interest. The good thing about open source is that we can fork it and even release a fork to the extension store. In fact, gdh1995 has a fork/rewrite. I appreciate their efforts to fix the original project, and I'm curious how stable Vimium-C is.

[gdh1995]

not as stable as Vimium; not all functions have been merged. And, ~~unfortunately, it was unpublished by Google yesterday, because of unsuitable description/snapshot. I'll try to complain in these days~~. Vimium C can still be found in recent days, but any new version of it is / will being reviewed by Google, although I don't really know how to pass it successfully.

…

---Original--- From: "Sebastian Sangervasi"<[email protected]> Date: Fri, May 10, 2019 04:34 AM To: "philc/vimium"<[email protected]>; Cc: "Mention"<[email protected]>;"Dahan Gong"<[email protected]>; Subject: Re: [philc/vimium] fix broken copy, paste and showFindMode since Chrome 74 (#3277) innaterebel was a bit sassy, but it is likely the maintainers have lost interest. The good thing about open source is that we can fork it and even release a fork to the extension store. In fact, gdh1995 has a fork/rewrite. I appreciate their efforts to fix the original project, and I'm curious how stable Vimium-C is. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[doingnothing]

Could you help me?
Failed to load extension
File
~/Downloads/vimium-fix-focus-of-hud
Error
Could not load background script 'lib/utils.js'.
Could not load manifest.

[gdh1995]

vimium_1.64.5-merge-3277.zip

@doingnothing This is my compiled version of this PR.

If you want to use Vimium from source code, then you need to "compile" Vimium, and you need to install Node.js, CoffeeScript, and run cake build. Refer: https://github.com/philc/vimium/blob/master/CONTRIBUTING.md#installing-from-source

[doingnothing]

vimium_1.64.5-merge-3277.zip

@doingnothing This is my compiled version of this PR.

If you want to use Vimium from source code, then you need to "compile" Vimium, and you need to install Node.js, CoffeeScript, and run cake build. Refer: https://github.com/philc/vimium/blob/master/CONTRIBUTING.md#installing-from-source

I just found your extension on Chrome Web Store.
https://chrome.google.com/webstore/detail/vimium-c/hfjbmagddngcpeloejdejnfgbamkjaeg?hl=en-US
And successfully installed it.
It seems to work perfectly.
Thank you very much!

[elongl]

@gdh1995
Should I replace my Vimium with your Vimium C?
Would you keep maintaining it?

[gdh1995]

only a few features like across-frame hints, filtered hints, and search word completion are no in Vimium C. So, if you don't use them, you may replace vimium with vimium c. I'll continue to develop and maintain vimium c for at least 2 more years, and if no accidents, for a decade ^_^.

…

---Original--- From: "Elon Gliksberg"<[email protected]> Date: Sat, May 11, 2019 01:24 AM To: "philc/vimium"<[email protected]>; Cc: "Mention"<[email protected]>;"Dahan Gong"<[email protected]>; Subject: Re: [philc/vimium] fix broken copy, paste and showFindMode since Chrome 74 (#3277) @gdh1995 Should I replace my Vimium with your Vimium C? Would you keep maintaining it? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub, or mute the thread.

[innaterebel]

So why doesn't someone else fork the project and just basically take ownership of it?

So why did you ask this when there has always been one and is mentioned in this thread (and many other places in this issue tracker)...

Someone who actually cares about Vimium.

Don't get me wrong, I fully appreciate all developers who has been participated in this project for their honorably wasting of their time. I was not complaining about them for anything.
I am not a native speaker, not familiar with the actual impression and extra nuance a word may imply to the native speaker, some unintentional strong words may be used. I may sound sarcastic or something, if any, is not toward the developers.
I just wanted to explain the actually long-standing circumstance as more and more people could not see it.

It is less than obvious to people who are new to Vimium (moreover, new to this repository pages), but the developing power of Vimium is in fact currently in its lowest, since just before Firefox officially dropped the support of their versatile legacy add-ons, which introduced a whole lot of demand to the simplism Vimium.
With the departure of the last contributors, and the remain only committer apparently ran out his time or motivation to do major updates to the project. In the past couple months the only thing that was gain is the kind and helpful gdh1995.

I filed this focus/copy issue one month before it reached the stable version of Chrome, worrying the only committer would go completely offline anytime soon (the owner has been almost offline for years), but it was still too late.
It's a pity the developers gone without saying hi, but that's about it, the nature of open source volunteering, again not blaming any of them. We don't know any fact about their missing after all. The only thing we can know through a few clicks in this repository is the stressful demand to the project. If you are unaware, this is the story.

[philc]

Hey folks, bandwidth from the committers is low at the moment, but not zero. Rest assured we'll get this fix into the chrome store release soon. Stephen or I will follow up on this issue in the next 48 hours.

…

On Fri, May 10, 2019 at 3:58 PM, innaterebel < ***@***.*** > wrote: > > > So why doesn't someone else fork the project and just basically take > ownership of it? > > So why did you ask this when there has always been one and is mentioned in this thread (and many other places in this issue tracker)... > > > Someone who actually cares about Vimium. > > Don't get me wrong, I fully appreciate all developers who has been participated in this project for their honorably wasting of their time. I was not complaining about them for anything. I am not a native speaker, not familiar with the actual impression and extra nuance a word may imply to the native speaker, some unintentional strong words may be used. I may sound sarcastic or something, if any, is not toward the developers. I just wanted to explain the actually long-standing circumstance as more and more people could not see it. It is less than obvious to people who are new to Vimium (moreover, new to this repository pages), but the developing power of Vimium is in fact currently in its lowest, since just before Firefox officially dropped the support of their versatile legacy add-ons, which introduced a whole lot of demand to the simplism Vimium. With the departure of the last contributors, and the remain only committer apparently ran out his time or motivation to do major updates to the project. In the past couple months the only thing that was gain is the kind and helpful gdh1995. I filed this focus/copy issue one month before it reached the stable version of Chrome, worrying the only committer would go completely offline anytime soon (the owner has been almost offline for years), but it was still too late. It's a pity the developers gone without saying hi, but that's about it, the nature of open source volunteering, again not blaming any of them. We don't know any fact about their missing after all. The only thing we can know though a few clicks in this repository is the stressful demand to the project. If you cannot see it, this is the story. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub ( #3277 (comment) ) , or mute the thread ( https://github.com/notifications/unsubscribe-auth/AAACDFWNOFHFACDCK24CAMLPUX4YJANCNFSM4HAUFNYQ ).

[athakral]

Ye ye he is back :)

[philc]

@gdh1995 great work on this. This problem was tricky and the fix is thoroughly researched and explained. I'll merge and cut a new release.

[philc]

This is now on the chrome store as v1.64.6 and should roll out in the next few hours.

[adriangonzalez]

/ is working for find, but n-Next and p-Previous aren't working for me?

[adriangonzalez]

Working now.

[gdh1995]

For who are interested in the browser code behind, here's what I find in the source code of Chromium:

https://github.com/gdh1995/vimium-c/blob/419e0c3067606f32f94779326de60f84ddbbf507/types/vimium_c.d.ts#L784-L801