Does this happen both with --enable-features=Polymer2 and --disable-features=Polymer2 ?

Can you bisect with tools/bisect-builds.py ?

Correction:
--enable-features=WebUIPolymer2 and --disable-features=WebUIPolymer2

I'm dealing with another P1 right now so I don't have time to experiment but it was happening from HEAD when I ran it. I don't think I set either of those flags.

It happened on a colleague's Chromebook as well so I think it should be easy to repro.

Seems to be a problem regardless of the flag. I'll take a look.

There is a two-way binding that is misbehaving.

<iron-icon icon="[[getIcon_(bluetoothToggleState_)]]"></iron-icon>
<cr-toggle id="enableBluetooth" checked="{{bluetoothToggleState_}}"
</cr-toggle>

When cr-toggle changes states, getIcon_ gets called without bluetoothToggleState_ actually being updated so we return the wrong icon. There haven't been any changes to the Bluetooth code in a while so I'm suspecting something changed at a lower layer.

This worked on m70 I'll ask for bisect to try to narrow down what changed.

Repro steps for the bisect:

1. Open chrome://settings
2. Scroll down to bluetooth
3. Click the Toggle

Expected:
If it changed to Powered On, the icon should be like [1]. If it changed to Powered Off, the icon should be like [2].

Actual:
If it changes to Powered On, the icon is like [2], and if it changes to Powered Off, the icon is like [1].

[1] https://www.materialui.co/icon/bluetooth
[2] https://www.materialui.co/icon/bluetooth-disabled

Indeed this a race condition between when Polymer calls observers (getIcon_() in this case), and when it updates two-way bound variables. An easy way to see this is to add a 
console.log() statement  at [1] within the getIcon_() observer comparing the observer's parameter (which has the correct value) to this.bluetoothToggleState_ which has not been updated yet (has the previous value).

The following fixes it https://chromium-review.googlesource.com/c/chromium/src/+/1289410.

@ortuno: Do you already have a fix out? If not I can send this one for review.


[1] https://cs.chromium.org/chromium/src/chrome/browser/resources/settings/bluetooth_page/bluetooth_page.js?type=cs&g=0&l=168

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/861cf56d6f6cbf4dff06e34139f1e2a97132aae0

commit 861cf56d6f6cbf4dff06e34139f1e2a97132aae0
Author: dpapad <dpapad@chromium.org>
Date: Thu Oct 18 22:47:23 2018

ChromeOS Settings: Fix bluetooth icon updating.

There is a race condition between when Polymer calls observers and when it
updates a two-way bound variable in the parent element.

Bypass the isuse by using the observers parameter instead.

Bug:  896509 
Change-Id: I63b13c858b62214d7973af641f67c659de9ea674
Reviewed-on: https://chromium-review.googlesource.com/c/1289410
Reviewed-by: Giovanni Ortuño Urquidi <ortuno@chromium.org>
Commit-Queue: Demetrios Papadopoulos <dpapad@chromium.org>
Cr-Commit-Position: refs/heads/master@{#600951}
[modify] https://crrev.com/861cf56d6f6cbf4dff06e34139f1e2a97132aae0/chrome/browser/resources/settings/bluetooth_page/bluetooth_page.js

@ortuno: Do you think this should/could be merged to M71?

Yes, since this is a regression. I'll wait for canary to verify and then request a merge.