@jonross could you take a look.
I suspect https://codereview.chromium.org/2770723002, but not certain.

When in tablet mode, did you have the lid fully rotated?
How were you holding the device when you signed out? (Landscape vs Portrait)

fully rotated. Portrait.
(sent from phone)

2017年3月29日(水) 23:12 jonross via monorail <
monorail+v2.2824921257@chromium.org>:

I see the issue on 58, there seems to be a delay before the sensors are able to get a reading that it is in tablet mode.

I don't think that the patch cited in #1 is in 58. That chance first landed in 59.0.3041.0 and I didn't merge it into 58 myself.

I'll take a look to see what's happening

+gwendal@ & aaboagye@

Hey,

I rolled a build with debug logs in the chrome side. After the user logs out chrome restarts. During this time we are receiving empty values from the acceleromters. The tablet mode switch is similarly reporting as off.

There is a several second delay before chrome begins receiving valid input from the accelerometers. And during this time we begin rapidly receiving tablet mode on/off signals.

Any thoughts about recent changes?

 Issue 706185  has been merged into this issue.

I'm not able to reproduce this on my kevin. What firmwares are you all using? I tried with a guest account instead of actually signing in though.

1. Browse as Guest
2. Rotate screen to tablet mode.
3. Rotate device to portrait orientation.
4. Tap "Exit guest" from the shelf.
5. Login screen is still in portrait orientation and virtual keyboard works.

My version info is below:

CHROME_VERSION: 58.0.3029.39
CHROMEOS_RELEASE_VERSION: 9334.23.0

bios version: Google_Kevin.8785.163.0
EC version: kevin_v1.10.174-928e3d0

(You can find these from chrome://system)


Also, if you can reliably reproduce, can you send feedback and mention my username in the feedback report. We can try and take a look the logs.

This is easier to repro when the device is in portrait and nearly vertical (I was resting the edge on a desk)

I'm having about 70% repro rate, both using an account and guest.

Chrome: 59.0.3051.3
Chrome OS: 9403.0.0
Bios: Google_Kevin.8785.58.2016_10_17_1858
EC: kevin_v1.10.71-d21368f

I filled a feedback report and mentioned both of us and this bug in it.

jonross@, your firmware looks out of date. Can you run `chromeos-firmwareupdate -m autoupdate` and see if you can still repro?

The upgrade brought me to 1.10.174 as mentioned in #7. I'm surprised I was behind, I flashed the latest image when I went to test. Are we not incorporating the correct EC version in builds?

Anyways this EC helps alot, however it does not fix the issue. I was able to get a repro 1/10 times. Feedback filed.

Thanks for filing feedback, I think I might have a better understanding of the issue. Taking a look at your feedback log, and especially since you mention it occurs when it's nearly vertical, I think this is kind of an edge case and is to be expected.

When the hinge aligns too closely with gravity (a.k.a portrait landscape perpendicular to the ground), we don't make any decisions because the angle is unreliable. Therefore, in that position, no changes to the window manager are made.

Since Chrome restarts on user logout, it loses the previous state that it was in (touchview + portrait orientation) and since we're in the "unreliable state" we don't make any correction to put us back.

I'd expect if you tilted the device forward or back, it would correct itself.

Maybe Chrome could save the previous state, and use that across a restart until it forms reliable calculations of its own. However, I'm not sure if that would even be desired.

Regarding the old EC versions, as far as I'm aware, installing a test image will not automatically upgrade your firmware. That only happens when receiving updates OTA. (I believe during the ChromeOSPostInstall step)

I do see some interesting transitions if you kept the device vertical the whole time during your test. It looks like you were resting it on the right hand edge. 

EC log:

[2263.884 TM 1 342 deg] # screen folded back
...snip...
[2280.592 Button 'Volume Up' was pressed]
[2280.593 buttons: 2]
[2280.853 Button 'Volume Up' was released]
[2280.853 buttons: 0]
[2284.137 Button 'Volume Down' was pressed]
[2284.138 buttons: 4]
[2284.214 Button 'Volume Up' was pressed]
[2284.215 buttons: 6]
[2284.226 TM 0 123 deg] 
[2285.029 Button 'Volume Up' was released]
[2285.030 buttons: 4]
[2285.098 Button 'Volume Down' was released]
[2285.099 buttons: 0]
[2285.711 Button 'Volume Up' was pressed]
[2285.711 buttons: 2]
[2286.435 Button 'Volume Up' was released]
[2286.435 buttons: 0]
[2286.740 Button 'Volume Up' was pressed]
[2286.740 buttons: 2]
[2287.781 Button 'Volume Up' was released]
[2287.782 buttons: 0]
[2293.451 KB poll]
[2294.350 TM 1 325 deg]
[2294.552 KB wait]
[2295.209 KB poll]
[2296.379 KB wait]
[2330.772 KB poll]
[2332.058 TM 0 128 deg]
[2332.397 KB wait]
[2333.875 KB poll]
[2337.576 KB wait]
[2347.463 KB poll]
[2348.631 KB wait]
[2376.701 KB poll]
[2377.771 KB wait]
[2378.799 TM 1 355 deg]
[2393.730 TM 0 131 deg]

This test was done at nearly vertical on the right hand side. Hence the volume keys.

Similar to the EC calcing the SW_TABLET_MODE switch, Chrome can't determine the hinge angle when nearly vertical.

Thanks for the clarification about the EC updates on direct instal..

Since, with the correct EC version, accelerometer values are being sent upon restart, Chrome is able to correctly determine the state of tablet mode upon logout. Except in the noted literal edge-case.

I plan to close this issue as Won't Fix (not reproducible)

There is already issue 703678 tracking a request to maintain tablet orientation between login screen and user state. However, since Chrome restarts, this will require a way to safe state.

I always have this issue after signout. 
1. Login 
2. Lie the device on the desk as laptop mode, and sign out.
3. Virtual keyboard does not appear on sign in screen.

I paste the chrome:version below. I updated the chromeos for development, but it did not resolve the issue.

Google Chrome	59.0.3048.0 (Official Build) unknown (32-bit)
Revision	0
Platform	9391.0.0 (Official Build) dev-channel kevin test
Firmware Version	Google_Kevin.8785.163.0
Customization ID	SAMSUNG-KEVIN1
ARC	3836840
JavaScript	V8 5.9.66
Flash	25.0.0.127 /opt/google/chrome/pepper/libpepflashplayer.so
User Agent	Mozilla/5.0 (X11; CrOS aarch64 9391.0.0) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/59.0.3048.0 Safari/537.36

oka@,

Can you clarify your steps? For step 2) did you mean "lie the device on the desk as *tablet mode*, and sign out"?

If so, I think that's the same issue that jonross@ was saying.

> "There is already issue 703678 tracking a request to maintain tablet orientation between login screen and user state. However, since Chrome restarts, this will require a way to safe state."

When you reproduce this, can you file feedback and mention my username? I can take a look at the logs tomorrow.

Hi aaboagye@.

the desk as *tablet mode*, and sign out"?
Yes.

I filed feedback mentioning aaboagye@. Hope it helps. Thank you.

oka@ can you also check the EC version?

Go to: chrome://system
Scroll down to ec_info

If it does not say kevin_v1.10.174-928e3d0 then you are on an older version of the EC with some issues. This component is not updated if you manually flash builds.

As for step 2) mentioned in #13 that is tracked by issue 703678

oka@, found your feedback report. It seems like you're running the latest EC FW, however the underlying issue is the same as issue 703678.

jonross@, you can probably dupe this if you want.

Now that the EC firmware has been confirmed as resolved, all that is left to address is already being tracked in issue 703678