I do not find how to make this report private and mark it as potential security issue.

I will create a simple example to show you, for investigating on the whole game would be uselessly too complicated.

For the moment I don't get to replicate it in a simple example.

The game's code is huge and compressed so unreadable. This is why I try to write a simple, clear and easy to read, example. For some reason the bug does not happen in it, there must be something done in the game that is needed to be done in order for the bug to happen. I am still searching.

Here is a video showing :
http://diegotests.altervista.org/bugs/20180122_bg_flickering/MVI_2305_2.mp4
Here is a copy of my game :
http://diegotests.altervista.org/html5_games/8bpkwr_20180122/index.php

I need to point something out. When I come back to the title screen at the end of the video, you can see the previous screen, the table of save games, flashing.
Between the two screens, the canvas this table was drawn on was cleared, and the picture it was made of was unloaded from memory.
So it is a 'screenshot' of a previous state of the canvas.
Such 'screenshot' sometimes comes back after I reload the page. This is why I think it is a security issue. It could be another user's confidential datas, it should not stay in memory like this and appear.

( By the way, while searching if the bug was already reported I read about a money reward for reporting security related bugs. If it is confirmed that this bug is a security bug like I think it is, is there a chance I get a reward ? )

@didi.tablette: Thanks for the report!!

Could you please help us with a sample file where you're seeing this issue and also provide details of your device, that would help us in triaging the issue.

Thanks!!

I found out that in order for the bug to happen it needs several things :
- There first must be a globalCompositeOperation operation on a canvas which is drawn on the still canvas.
- Something must be drawn on the canvas on top on the still canvas too.

I got to make a very simple example that makes the bug happen.
Here :
http://diegotests.altervista.org/bugs/20180122_bg_flickering/1.htm

In Chrome on a computer, everything is ok.
On my tablet it is flashing everytime it stops drawing on the canvas.
Video :
http://diegotests.altervista.org/bugs/20180122_bg_flickering/MVI_2308_2.mp4

My tablet :  iFive Mini 3 (aka Haier mini Pad 781) , Android 4.4.2 , RK3188 SOC, 1GB RAM, 1240x768
Unfortunatly I do not have any other android device to try on at the moment.

Thank you for providing more feedback. Adding requester "sandeepkumars@chromium.org" to the cc list and removing "Needs-Feedback" label.

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

Tested the issue using #64.0.3282.137 on Sony 4.4.4; C6902 Build/14.4.A.0.108 and could not see any flickering by navigating to http://diegotests.altervista.org/bugs/20180122_bg_flickering/1.htm

Unable to check the issue on iFive Mini 3 as we don't have mentioned device with us.

Hence removing Needs-TestConfirmation Needs-Bisect labels for now, Please add it back if required.

@didi: Could you please update Chrome to #64.0.3282.137 and check if you still face the issue?

Thanks!!!

Ping @didi: Can you still reproduce this issue with a current version of Chrome?

(oops, so sorry I didn't see that post in february)
Yes, it is still happening on Chrome 65.0.3325.109

Re-triaging as a compositing bug.

I am unable to repro with the hardware at my disposal.
This looks like a bug with layer visibility -> flagging the Compositor team.

Anything that only repros on a single device is most likely GPU, unless it triggers the low-end device systems. I don't think this one does.

At present, there's no indication that this is a security issue. There would be a security issue here only in the event that a "polluted" canvas were readable via JavaScript (e.g. if a call to canvas.toDataURL() returned the lingering pixels). If we were able to come up with a repro that resulted in unexpected data in the PNG returned by toDataURL, *that* would potentially be a security issue.

Based on the text here, that's not what's happening, and instead what we have here is solely a visual corruption that cannot be abused by a malicious site.

GPU is "Mali-400 MP4", in Rockchip RK3188 SOC.
Talking about GPU and composition, this reminds me that while working on a game of mine using Crosswalk, starting it with "xwalk --ignore-gpu-blacklist --show-fps-counter --ui-show-fps-counter" commands, it said that some hardware composition acceleration feature (sorry I do not remember the right name) could not be used on this tablet because the GPU is not compatible. And while this game is sluggish in Crosswalk, it is a LOT faster in Chrome. And Crosswalk is basically just an old version of Chrome. So maybe Chrome is now forcing this feature on, ignoring that this GPU can't handle it correctly. How can I check ?

Still happening as of v71.0.3578.99.
And I just saw the same bug on another tablet, a Chuwi VX3, SOC is Mediatek MT6592, GPU is Mali450.