Internal Bug :b/63745109

DUP of  crbug.com/741306 ?

Reopening; I ran bisection and found that this is a bug distinct from  crbug.com/741306 .
(see https://bugs.chromium.org/p/chromium/issues/detail?id=741306#c17 for the timeline.)


Bisection pointed out https://chromium-review.googlesource.com/c/552719/

Shall we revert?

Confirmed that ToT + local revert fixes the issue. Revert SGTM. I'll create a CL

Revert landed: https://chromium-review.googlesource.com/c/575294

I did a little investigation and found out below fact.

1. The CL[1] + ChromeOS 9608 doesn't have the back window issue.
2. The CL[1] + ChromeOS 9708 has the back window issue.
3. The CL[1] works fine on Workstation with wayland clients.
4. I also tried modify CL[1] to replace texture quads with solid color quads for all exo::Surface. And exo shows all surfaces as rectangles with random colors on screen for ARC++ apps with ChromeOS 9608. But for ChromeOS 9708, the Window is still black.


[1] https://chromium-review.googlesource.com/c/552719/

Base on those facts, I believe some changes in ChromeOS or ARC are related to this issue. The change should be between 9608 and 9708.

Do we expect the revert to fix per #8 or still an open issue per #9?

Assume this would have made it for the late build on 7/17?

Revert #8 fixed the issue.

61.0.3160.0 (r487322) does NOT include the revert (r487392):
https://chromium.googlesource.com/chromium/src/+/61.0.3160.0
https://chromium.googlesource.com/chromium/src/+/4f19a78b4e3f0a3242bec6938ae55f49f74fa210

if the fix is necessary for the branch, the CL has to be cherry-picked manually by someone (sorry, I myself cannot help since I'm off the office now.)

I tested ChromeOS 9609 & 9610. And found out the R61-9610 is the first ChromeOS image which doesn't work with my CL.

With which device are you testing?

For instance Samus/Cyan/Minnie switched from M to N at that point and that'll surely affect a lot.
https://crosland.corp.google.com/log/9609.0.0..9610.0.0

I am using veyron_minnie for development. Can I manually switch back to M?

I tested exo with several wayland applications on device with ChromeOS 9610. All of them works fine. So the problem must related to changes in ARC++.

+Bernie may know how to do that (recent minnie + M).

Or, "veyron_minnie-cheets" board/images are using N even on revisions older than 9609. You can test 9608+N for instance by using the board.

If you build with USE="${USE} -android-container-nyc android-container" you should be able to get the old M container, I put up a quick CL at https://chrome-internal-review.googlesource.com/#/c/414189/ that should do this.

Note that M is almost entirely deprecated at this point, only veyron_fievel and veyron_tiger are left with it in M61, so it should still work but is more liable to break unnoticed than N.

I found the problem is because ARC++ app window is covered by a black box. See the video https://photos.app.goo.gl/F6VbFpYqCdxmdt9j1 

Any idea about where the black box come from?

+revema,+lpique, +domlaskowski, +oshima on #18

Has this been tested with the latest M61 dev release at 9756.1.0, 61.0.3159.8?  I was under the impression a fix had been implemented?

Has this been tested with the latest M61 dev release at 9756.1.0, 61.0.3159.8?  I was under the impression a fix had been implemented?

Verified that issue is fixed on M61 dev release 9756.1.0, 61.0.3159.8

Can we remove the dev-blocker tag since it's verified fixed and it'd block dev releases?  Thanks

The reverted CL has been relanded with https://chromium-review.googlesource.com/c/582932. The problem has been fixed.

Verified on Chrome OS 9840.0.0, 62.0.3176.0 dev build as well. Closing this bug.