CL: https://chromium-review.googlesource.com/c/563143/

Applying RBS since this is a user-visible regression (though a minor one).

Timing is a bit tight. This fix requires two CLs to land:

1. https://chromium-review.googlesource.com/c/564869/
2. https://chromium-review.googlesource.com/c/563143/

The first is to fix a bug that was introduced since r473830 that prevents that CL from being reverted.

If it's not possible to merge back to M60 at this late stage, we'll just have to have the bug in one release. Note that the first CL does not need to be merged to M60.

Another minimal fix here is to add a "select none" call back just before the "set display offset to 0" that was added to OmniboxViewViews::OnBlur().  That would hopefully be minimal enough not to touch anything else and thus be suitable for immediate landing/merging.

#4 that is what my CL does:
https://chromium-review.googlesource.com/c/563143/7/chrome/browser/ui/views/omnibox/omnibox_view_views.cc

The problem here is that r483795 has introduced some (IMHO invalid) logic that makes ChromeVox read out the name of a textfield if its text-selection changes, even if it isn't focused. So adding that "select none" back as you suggest actually breaks ChromeVox behaviour and there is a browser test (SpokenFeedbackTest.FocusToolbar) that fails.

I'm trying to fix this in https://crrev.com/c/564869 but encountering some resistance. Since r483795 isn't in M60, fixing the ChromeVox issue is only needed to land this in M61 (not to merge). So the other approach I am considering is landing https://crrev.com/c/563143 with disabling SpokenFeedbackTest.FocusToolbar, merging it, then re-enabling the disabled test as part of a later fix for ChromeVox.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/5014480e8bf212ec81f8c8cd298fb73e779eb092

commit 5014480e8bf212ec81f8c8cd298fb73e779eb092
Author: Matt Giuca <mgiuca@chromium.org>
Date: Tue Jul 18 04:34:44 2017

Views Omnibox: Fixed text being selected even after blur event.

This regressed in r473830 (blurring the Omnibox then mouse-overing
resulted in an arrow cursor, not an I-beam). Now it properly deselects,
so the cursor shows as an I-beam again.

Bug:  740008 
Change-Id: Ice2b4fdb21acf642bdf928c07c40beec5639274f
Reviewed-on: https://chromium-review.googlesource.com/563143
Commit-Queue: Matt Giuca <mgiuca@chromium.org>
Reviewed-by: Justin Donnelly <jdonnelly@chromium.org>
Cr-Commit-Position: refs/heads/master@{#487386}
[modify] https://crrev.com/5014480e8bf212ec81f8c8cd298fb73e779eb092/chrome/browser/ui/views/omnibox/omnibox_view_views.cc
[modify] https://crrev.com/5014480e8bf212ec81f8c8cd298fb73e779eb092/chrome/browser/ui/views/omnibox/omnibox_view_views_browsertest.cc
[modify] https://crrev.com/5014480e8bf212ec81f8c8cd298fb73e779eb092/chrome/browser/ui/views/omnibox/omnibox_view_views_unittest.cc

Should be fixed now.

So I'm requesting a merge to 60 (because it's a regression). I realise this is very late so it's fine if we decide not to merge. It's a very small regression (wrong mouse cursor is shown in Omnibox in certain cases). Note that:

- This is a one-line fix (modulo tests):
  https://chromium.googlesource.com/chromium/src/+/5014480e8bf212ec81f8c8cd298fb73e779eb092%5E%21/#F0
- r487386 will not properly apply between r483795 and r487350 (in that range is a ChromeVox bug that prevents this CL from cleanly applying). However, that won't be an issue as M60 branched much earlier than that range. TL;DR: r487350 does NOT need to be merged to M60.
- Have not yet tested on Canary, but I will at the earliest convenience.

This bug requires manual review: We are only 6 days from stable.
Please contact the milestone owner if you have questions.
Owners: amineer@(Android), cmasso@(iOS), josafat@(ChromeOS), bustamante@(Desktop)

For more details visit https://www.chromium.org/issue-tracking/autotriage - Your friendly Sheriffbot

+ bustamante@ for M60-Merge review.

I'd like to wait until M61 as it's very late and this isn't a particularly rejected.  Rejecting for merge and moving to 61.

> this isn't a particularly rejected.

I assume you mean "this isn't a particularly bad bug"?

Ack.

 Issue 760031  has been merged into this issue.