Tested the issue in Android and able to reproduce the issue. 

Steps Followed:
1. Navigated to https://snipg.github.io/
2. Observed white screen , on refreshing page it loaded succesfully.

Chrome versions tested:
67.0.3396.87, 68.0.3440.70 , 70.0.3508.0(Latest canary)

OS:
Android 7.0.0

Android Devices:
Samsung galaxy J7, Mali GPU

Issue is seen in latest stable but not seen in latest dev #69.0.3497.9 and latest canary #70.0.3508.0. Hence providing reverse bisect info.

Using the per-revision bisect providing the bisect results,
Last Bad Build - 69.0.3489.0
First Good Build - 69.0.3491.0 

CL: https://chromium.googlesource.com/chromium/src/+/4d1d8967490fa5c8bc7689622ca2f3820aa8b376

@ jinsukkim: Could you please look into the issue, pardon me if it has nothing to do with your changes and if possible please assign it to owner concerned. Adding RB-Stable for M-68, Please remove if this not the case. Please merge fix to M-68 if this is safe.

Please navigate to below link for log's  --
go/chrome-androidlogs/868679

Thanks!

Just for clarification - does the problem happen on all the following version?

> Chrome versions tested:
> 67.0.3396.87, 68.0.3440.70 , 70.0.3508.0(Latest canary)

My CL landed at 69.0.3491.0, so cannot have caused the bug for 67/68.

Unassigning myself - let me know if more bisects repeatedly point to mine. I don't have a device with Mali GPU so it's not possible for me to reproduce the issue locally.

 chelamcherla@ can you follow up here please?

Here's the log for the *reverse* bisect range - i.e. apparently it was fixed in this range:
https://chromium.googlesource.com/chromium/src/+log/69.0.3489.0..69.0.3491.0?n=10000
(I'm confused, though, at comment #3 - shouldn't per-revision bisect have given a smaller range?)

Also note that if the issue is flaky, the bisect range could be completely wrong.

At a glance, there are plenty of things in this range that could have had an effect.

reporter: could you please confirm whether the issue has been fixed in Canary (70) or Dev (69) - as indicated by the bisect? If so, maybe there is nothing to do here. (Issue 865946 could be separate.)

I used: Chrome Canary 70.0.3511.0, Chrome Dev 69.0.3497.24, chrome 67.0.3396.87

Phone: Coolpad e502 - Mali-T720MP2. Canary and Dev showed website first time without problems. While Chrome 67 fails and needs reload. No errors on consol.

But i got secound phone to test, and results are different.

Phone Samsung A3 2017 - Mali -T830 MP1 . All 3 browser fail, and need reloading to display website.
While Samsung Internet 7.2.10.33 displays website everytime.

I don't have more phones to test with, but it seems some bad behavior is still there.
As i don't own thous phones, so iam not sure if they are on latest system update etc, if it even matters.

Some extra info: I saw that Comment 3 phone Samsung galaxy J7, if its 2017 version should have same GPU as Samsung A3 2017(Model: SM-A320FL), which still shows bug on all browser versions, so maybe android version or some other settings affects it to. 

Samsung A3 2017 Has android 8.0.0, in developer tools -> webview  settings default android System webview is disabled and it uses chrome, changing it to canarly or dev, does not fix problem.
I don't know how to enable default system webview on A3, to test with that.
If webview maybe only affects apps, then this does not even matter.

Coolpad uses Default Android system webview 67.0.3396.87 same as chrome version. 

hexerdomix@gmail.com: can you share about:gpu page content on phones/chrome where WebGL2 doesn't work?

Adding zip folder with constains about:gpu info from chrome 68, dev and latest canarly build. 

Deleted: ChromeCasual_Dev_Canarly.zip 17.0 KB

ChromeCasual_Dev_Canarly.zip 17.0 KB Download

Assigning to zmo@ to take a look.

Tested on a couple of phones. The Samsung Galaxy S7-2 (SM-G930F) with Mali-T880 running NRD90M.G930FXXU1DQL7 wouldn't run the Chrome official per-revision builds; they all crashed.

The Galaxy A7 2017 (SM-A720F) with Mali-T830 running NRD90M.A720FXXU2BQI9 ran those builds successfully. Chrome Beta (69.0.3497.41 32-bit) runs the test case https://snipg.github.io successfully the first time, as does Dev Channel and Canary. Ran a per-revision bisect as follows:
python bisect_builds.py --use-local-cache -o -a arm -g 582755 -b 561733

Result was:

You are probably looking for a change made after 574815 (known bad), but no later than 574816 (first known good).
CHANGELOG URL:
The script might not always return single CL as suspect as some perf builds might get missing due to failure.
  https://chromium.googlesource.com/chromium/src/+log/4d1d8967490fa5c8bc7689622ca2f3820aa8b376..1c3333c5acf70c841a66feba15273b1dc7d76da0

which was jdarpinian's:
https://chromium.googlesource.com/chromium/src/+log/4d1d8967490fa5c8bc7689622ca2f3820aa8b376..1c3333c5acf70c841a66feba15273b1dc7d76da0
https://chromium.googlesource.com/chromium/src/+/1c3333c5acf70c841a66feba15273b1dc7d76da0

Submitter, this issue is already fixed in Chrome 69, which is currently in beta and slated to ship soon. Closing as WontFix.

Chrome 69 indeed fixes something and with one phone example issue does not occur. But it does not fix it on other phone and i don`t have more phones to test with. Issue like this could need larger example range for testing.

I saw you try to find fix for Issue 865946, and i bet if you find the cause and fix it, it most likely fix this issue as well, so i hope for best that you can find it.
To help you(kbr@) out a little, you need start construct 110.2 beta release to test stuff out.

Yesturday i tested issue with Samsung Galaxy S9 (G960F version), latest chrome Dev, and issue occured. Also it had chrome 67 on it, and when I opened link, ordinary you see white screen and when you reload it, you see sprite+text. But on 67(it occured once after ~10+ tests), the sprite was just blue, no text object or no original blue sprite with white square in it(could be some other issue).
Construct uses shaders to draw textures, so thats that. As you guys are experts maybe it rings some bells ... or not.


 

And can you guys maybe point me at  some link etc. Where i could info how i could myself analyze chrome internals, dev tools fall short with problems like these.