Unable to reproduce the issue on Chrome60.0.3112.0/ChromeOS 9592.0.0- reks

Yingbing, can you please take a look? Thanks.

Changed the keyboard lanuguage /OS language / text entry settings in ubuntu 14.04 to korean , but Korean IME is not reflecting . Tried in different machines and other languages are working fine.

@MTV Team -- Could anyone from MTV team , please check from your end , if you can triage the issue.

Thanks!

Hi Shu, This bug is refer Ubuntu OS not Chrome OS.
songsuk@, for step 2, Which IME? Google Input Tool or Ubuntu native IME?

I'm working on Korean IME with "Ubuntu native IME, ibus-hangul".

>>  for step 2, Which IME? Google Input Tool or Ubuntu native IME?

Will do Bisect on Linux

Bisect result :

CHANGELOG URL:
The script might not always return single CL as suspectas some perf builds might get missing due to failure.
  https://chromium.googlesource.com/chromium/src/+log/47564c7bf17d2fd560b4390ba714d19d4d9e2550..5fdb3237636167dfb535cd5ef8a535e5b9f8b662



@chongz,
Could you look at the issue. 

Was able to reproduce on Linux, 58.0.3029.110 (Official Build) (64-bit).
Please see attached video. 

songsuk@ Could you help verify if this is a regression or a long standing issue? Thanks!

Deleted: M58-Hangul-Bad.webm 86.7 KB

	M58-Hangul-Bad.webm 86.7 KB View Download

Actually I just realize I have 'Experimental Web Platform features' always enabled. Will look into it, thanks!

A simpler reproduce step:
1. Open Chrome M60 or M58/M59 with 'Experimental Web Platform features' enabled
2. Go to Google Doc
3. Press 'a'
4. Press space

Expected:
ㅁ_ (_ is space)

Actual:
ㅁㅁ_ (_ is space)

I've attached 2 images for the good event trace and the bad event trace, the differences are:
1. The bad trace has 'beforeinput' fired, however this shouldn't affect anything as this is a new event and no one should be using it;
2. The bad trace has 'data' field in 'input' event, but this shouldn't cause trouble as there is no 'data' field in the previous spec, and JS shouldn't be accessing it.


+jzoll@
Can I have some help on how Google Doc is using those events? Thanks!

Deleted: a-space-bad.png 115 KB

	a-space-bad.png 115 KB View Download

Deleted: a-space-good.png 96.4 KB

	a-space-good.png 96.4 KB View Download

So this is happening because we have two different systems conflicting here in Docs:
- the IME system which handles the composition events and ignores the input event which occurs afterwards.
- the system which handles beforeinput events of type 'insertText'.

The second system should only be being triggered in documents owned by Googlers. I will file an internal bug to fix this. Please let me know if you are seeing this on a production document not owned/created by a non-Googler account.

Re #15: That makes sense then, thanks for the info! 

songsuk@: Are you seeing this issue on non-Googler docs? I wasn't able to verify as Google Doc seems to be able to recognize my IP even if I'm signed-in with a different account.

Otherwise we can close this as WontFix since it has been tracked by an internal bug.

Actually, this doesn't seem as simple as I thought. When I try to reproduce the test from Comment 13 on both M58 and M61, all of my beforeInput events have input type 'insertCompositionText'.

Is it possible that M60 has an issue where it was showing insertText? 

I've attached the trace I get for both 58 and 61 when I Have Enable Experimental Web Platform features.

Deleted: chrome61.png 281 KB

	chrome61.png 281 KB View Download

Deleted: chrome58.png 286 KB

	chrome58.png 286 KB View Download

I'm seeing the issue on personal account, non-Google account. But, the account is connected to GoogleGuest while testing the docs. 

>>songsuk@: Are you seeing this issue on non-Googler docs? I wasn't able to verify as Google Doc seems to be able to recognize my IP even if I'm signed-in with a different account.

Sorry yes, I was looking at the wrong flag - the beforeinput 'insertText' events are being handled in production. Okay, I have filed an internal bug for this and will follow up. The 'beforeinput' event is always gated by that Enable Experimental Web Platform Features unless using M60, correct?

Re #19: Correct. It was gated by "Enable Experimental Web Platform Features" and the CL in #9 marked 'beforeinput' as enabled by default in M60.

FWIW 'beforeinput' was enabled by default in Safari 10.1 as well.

I've investigated and it turns out this is due to beforeinput 'insertText' events not being cancellable on Chrome, hence why we were double inserting. We will workaround on our end, but please CC me when this changes.

Suspecting this may caused by cl https://codereview.chromium.org/2872343003 which is being reverted https://codereview.chromium.org/2909063002.

The cl https://codereview.chromium.org/2916673002 fixes a dup event rewrite issue, maybe related to this.

Does this issue still repro on HEAD?

Re #22: Thanks for the information! I can still repro on HEAD so it should be a different issue.

The cause in #21 is tracked by b/62291255 and should be fixed after cl/157856712.

Close as b/62291255 has been marked as fixed.

[Auto-generated comment by a script] We noticed that this issue is targeted for M-60; it appears the fix may have landed after branch point, meaning a merge might be required. Please confirm if a merge is required here - if so add Merge-Request-60 label, otherwise remove Merge-TBD label. Thanks.

No merge required since it's a Google Doc side fix.