same for auto hide shelf behavior

Is this actually desired? The new behavior seems confusing to me, and I have some concerns about the robustness of the implementation. Currently we make a lot of assumptions that prefs associated with the shelf are associated with a single profile. If we plan to change that, we should do an audit and explicitly call out which prefs are associated with which profile, and make sure that will still work when Ash and Chrome are separate binaries.

Please also confirm that this has been run by UX.

This actually is a regression. I remembered in some old time shelf alignment is remembered and bound to the corresponding profile for switching multi-profile case.

Shelf position and auto-hide is also broken on logout for a single user. Should I open a separate issue? Just wondering if whatever broke this is related somehow.

+msw in case you know what this is.

willg76x - What do you mean it is broken on logout? On the sign-in screen the shelf should always be on the bottom.

This reproduces on ToT r441398 running on Linux desktop with --login-manager and two users.

My last big shelf change was https://codereview.chromium.org/2573703003 which landed the day after this bug was filed.

warx, this should be bisectable. Do you have time to do it?

1. Change position to left or right and/or enable auto-hide
2. Press Ctrl+Shift+QQ
3. Log back in

The shelf will no longer be in auto-hide or be aligned vertically when you sign back onto the desktop.

On Version 57.0.2970.0 canary (64-bit)

New wrinkle: It'll be in the proper position after the device wakes from suspend.

Going to open a separate issue, thanks for your time.

+msw, +jonross - Comment 6 sounds like the shelf alignment pref isn't being saved.

re comment #5:

I tried bisect. But for some unknown reason, the builds before revision 429576 cannot popup/run chromeos UI... The revision 429576 is still a bad revision.

Can someone else try the builds before 429576? This bug seems broken long time ago.

In M54 stable (from August 25) at r414607 is "good" (the shelf has different alignments for each account). I don't have an M55 stable device handy to test, but that will give you a data point from October 7.

willg76x, can you post a link to the new bug you filed? It's possible it's related.

msw, it's possible that switching to mojo for classic ash caused this behavior change, e.g. https://codereview.chromium.org/2391253004 - maybe there's a race between ChromeLauncherController::SetShelfAlignmentFromPrefs and the active user profile changing.

warx, per comment 2, you might want to check with PM/UX on this one. It seems kind of odd to me that a user would want different autohide/alignment settings on the same monitor.

54.0.2840.93 is fine.
55.0.2883.99 stable is also fine.

This is a regression, and wouldn't be surprised if there was some race now. The active profile does switch on user change. So it's possible that the setting is never written out if the mojo message is received late.

I can repro and have a local fix; this is indeed a regression from my CL.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/913296ee74616a4589b5fea9c5873ead36e0e321

commit 913296ee74616a4589b5fea9c5873ead36e0e321
Author: msw <msw@chromium.org>
Date: Fri Jan 06 20:59:50 2017

Use the correct profile for shelf prefs.

Used the attached profile; not the active profile.
Regressed in: https://codereview.chromium.org/2391253004

BUG= 675599 
TEST=Multi-login cros users get correct shelf prefs.
R=jamescook@chromium.org

Review-Url: https://codereview.chromium.org/2613303003
Cr-Commit-Position: refs/heads/master@{#442045}

[modify] https://crrev.com/913296ee74616a4589b5fea9c5873ead36e0e321/chrome/browser/ui/ash/launcher/chrome_launcher_controller.cc

This should be fixed by the above CL; please help verify, thanks!

 Issue 681242  has been merged into this issue.

 Issue 683400  has been merged into this issue.

per comment 13 CL, the regression is introduced in m56. msw@, do you know if a merge is safe to do a merge?

abodenha@, is this a RBS for M-56?

+gkihumba for M56

It should be safe to merge to 56, adding request.

I believe this : )

This bug requires manual review: We are only 9 days from stable.
Please contact the milestone owner if you have questions.
Owners: amineer@(clank), cmasso@(bling), gkihumba@(cros), bustamante@(desktop)

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

Thanks for fixing my labels, it's hard to get right on a phone.

Moving back to started, so this gets merge review attention

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/78cf9e379414559004f8446a83d41ad78fab71dd

commit 78cf9e379414559004f8446a83d41ad78fab71dd
Author: Michael Wasserman <msw@chromium.org>
Date: Mon Jan 23 18:01:54 2017

Merge to M-56 (branch 2924) Use the correct profile for shelf prefs.

Used the attached profile; not the active profile.
Regressed in: https://codereview.chromium.org/2391253004

BUG= 675599 
TEST=Multi-login cros users get correct shelf prefs.
R=jamescook@chromium.org

Review-Url: https://codereview.chromium.org/2613303003
Cr-Commit-Position: refs/heads/master@{#442045}
(cherry picked from commit 913296ee74616a4589b5fea9c5873ead36e0e321)

Review-Url: https://codereview.chromium.org/2647293003 .
Cr-Commit-Position: refs/branch-heads/2924@{#841}
Cr-Branched-From: 3a87aecc31cd1ffe751dd72c04e5a96a1fc8108a-refs/heads/master@{#433059}

[modify] https://crrev.com/78cf9e379414559004f8446a83d41ad78fab71dd/chrome/browser/ui/ash/launcher/chrome_launcher_controller.cc

Fixed; please help verify and lmk if any related issues require attention.