[solved] Hover broken with Chromium

cmtonkinson · 2014-09-26 16:38:54

The search results for my current issue are badly Google-washed with noise, so please accept my apologies if this is a re-post or Wiki-covered question.

Off-topic: Since when has the forum search here on bbs.archlinux.org been rate limited to one search per 30 seconds?

The System

~ $ uname -a
Linux host.example.com 3.16.3-1-ARCH #1 SMP PREEMPT Wed Sep 17 21:54:13 CEST 2014 x86_64 GNU/Linux

~ $ pacman -Q | grep 'xorg\|chromium\|shell'
chromium 37.0.2062.120-2
chromium-libpdf 1:36.0.1985.143-1
chromium-pepper-flash 1:15.0.0.152-1
gnome-shell 3.12.2-1
gnome-shell-extensions 3.12.2-1
gnome-shell-system-monitor-applet-git 660.1b632f9-1
xorg-bdftopcf 1.0.4-2
xorg-font-util 1.3.0-2
xorg-font-utils 7.6-4
xorg-fonts-alias 1.0.3-1
xorg-fonts-encodings 1.0.4-4
xorg-fonts-misc 1.0.1-3
xorg-mkfontdir 1.0.7-2
xorg-mkfontscale 1.1.1-1
xorg-server 1.16.1-1
xorg-server-common 1.16.1-1
xorg-setxkbmap 1.3.0-2
xorg-xauth 1.0.9-1
xorg-xhost 1.0.6-1
xorg-xinit 1.3.4-1
xorg-xkbcomp 1.2.4-2
xorg-xrdb 1.1.0-1
xorg-xset 1.2.3-1

The Issue
Chromium mouse hovering is ostensibly broken. When using Chromium, moving the mouse around the browser (both browser controls and page content) does not activate hover states under normal circumstances.

Example 1: Moving the cursor over the "tab close x" does not cause the x to gain it's normal red background.
Example 2: Moving the cursor over the address bar does not cause the cursor to change to "text" style.
Example 3: Moving the cursor over a bookmark icon does not cause the "button" style to appear.
Example 4: Moving the cursor over any HTML on a web page activates neither CSS hover effects (such as background colors, cursor style changes, or underlining of links) nor Javascript mouse events ("onenter", "onleave", et al.).

However, when actively scrolling the mouse wheel these effects work as expected (in addition to typcial scrolling of whatever content/window/panel the cursor is over at that moment).

What I've looked at
I've looked at pacman logs and have not been able to identify a cause. This had been broken for several weeks (sorry, I can't recall the exact timeframe) but resolved itself a day or two ago (again, I didn't note the exact time of the fix) but began misbehaving again this morning. Here are my pacman logs for the past few days:

~ $ sudo cat /var/log/pacman.log | grep upgraded
...
[2014-09-24 07:35] [PACMAN] upgraded firefox (32.0.1-1 -> 32.0.2-1)
[2014-09-24 07:35] [PACMAN] upgraded lib32-libgpg-error (1.13-1 -> 1.16-1)
[2014-09-24 09:07] [PACMAN] upgraded python2 (2.7.8-1 -> 2.7.8-2)
[2014-09-24 12:38] [PACMAN] upgraded bash (4.3.024-1 -> 4.3.024-2)
[2014-09-24 12:38] [PACMAN] upgraded ruby (2.1.2-1 -> 2.1.3-2)
[2014-09-24 20:48] [PACMAN] upgraded accountsservice (0.6.37-1 -> 0.6.38-1)
[2014-09-24 20:48] [PACMAN] upgraded ca-certificates-mozilla (3.17-3 -> 3.17.1-1)
[2014-09-24 20:48] [PACMAN] upgraded colord (1.2.2-1 -> 1.2.3-1)
[2014-09-24 20:48] [PACMAN] upgraded gstreamer (1.4.2-1 -> 1.4.3-1)
[2014-09-24 20:48] [PACMAN] upgraded gst-plugins-base-libs (1.4.2-1 -> 1.4.3-1)
[2014-09-24 20:48] [PACMAN] upgraded gst-plugins-bad (1.4.2-1 -> 1.4.3-1)
[2014-09-24 20:48] [PACMAN] upgraded gst-plugins-base (1.4.2-1 -> 1.4.3-1)
[2014-09-24 20:48] [PACMAN] upgraded gst-plugins-good (1.4.2-1 -> 1.4.3-1)
[2014-09-24 20:48] [PACMAN] upgraded nss (3.17-3 -> 3.17.1-1)
[2014-09-25 10:16] [PACMAN] upgraded firefox (32.0.2-1 -> 32.0.3-1)
[2014-09-25 10:16] [PACMAN] upgraded git (2.1.0-1 -> 2.1.1-1)
[2014-09-25 10:16] [PACMAN] upgraded libsodium (0.7.1-1 -> 1.0.0-1)
[2014-09-25 10:16] [PACMAN] upgraded thunderbird (31.1.1-1 -> 31.1.2-1)
[2014-09-25 12:42] [PACMAN] upgraded apr-util (1.5.3-4 -> 1.5.4-1)
[2014-09-25 16:23] [PACMAN] upgraded curl (7.38.0-1 -> 7.38.0-2)
[2014-09-26 07:43] [PACMAN] upgraded bash (4.3.024-2 -> 4.3.026-1)
[2014-09-26 07:43] [PACMAN] upgraded p11-kit (0.20.2-1 -> 0.20.7-3)
[2014-09-26 07:43] [PACMAN] upgraded gnutls (3.3.8-1 -> 3.3.8-2)
[2014-09-26 07:43] [PACMAN] upgraded openconnect (1:6.00-1 -> 1:6.00-3)
[2014-09-26 07:43] [PACMAN] upgraded pygtk (2.24.0-4 -> 2.24.0-5)
[2014-09-26 11:39] [PACMAN] upgraded nodejs (0.10.32-3 -> 0.10.32-4)
[2014-09-26 11:39] [PACMAN] upgraded qtwebkit (2.3.3-2 -> 2.3.4-1)

Do any of these packages look like they would be involved?

I don't know where to even begin looking at this problem. X? Gnome? Chromium? Something else? I don't even know where I should be looking to file a bug report. As far as my (again, very noisy) Googling goes, I can't seem to find anyone else experiencing this.

I have disabled all Chromium extensions, restarted Chromium, reset all chrome:///flags to default values, reinstalled Chromium, restarted X, even restarted the system. Nothing I've tried yet has had an an effect.

Starting the browser leaves this in the journal:

~ $ sudo journalctl -e
...
Sep 26 12:26:39 host.example.com chromium.desktop[27940]: [27940:27940:0926/122639:ERROR:desktop_window_tree_host_x11.cc(1478)] Not implemented reached in void views::DesktopWindowTreeHostX11::MapWindow(ui::WindowShowState)
Sep 26 12:26:39 host.example.com chromium.desktop[27940]: ATTENTION: default value of option force_s3tc_enable overridden by environment.
Sep 26 12:26:39 host.example.com chromium.desktop[27940]: [27975:27975:0926/122639:ERROR:sandbox_linux.cc(308)] InitializeSandbox() called with multiple threads in process gpu-process
Sep 26 12:26:39 host.example.com gnome-session[1527]: Window manager warning: Buggy client sent a _NET_ACTIVE_WINDOW message with a timestamp of 0 for 0x2e00001 (Untitled -)
Sep 26 12:26:39 host.example.com gnome-session[1527]: Window manager warning: meta_window_activate called by a pager with a 0 timestamp; the pager needs to be fixed.
Sep 26 12:26:39 host.example.com chromium.desktop[27940]: [27975:27975:0926/122639:ERROR:sync_control_vsync_provider.cc(60)] glXGetSyncValuesOML should not return TRUE with a media stream counter of 0.
...

Which seems to suggest (from "Window manager warning: Buggy client sent a _NET_ACTIVE_WINDOW") that the browser and DE aren't communicating on the state of the Chromium window? But again, is this Chromium, X, or Gnome? The Python update on the 24th might correspond to the "fixing" of the problem, and the pygtk update this morning could have been it re-breaking, but that's a total guess. Does GTK depend on Python at runtime, and if so, could this problem be there?

Cheers!

Last edited by cmtonkinson (2014-11-15 01:20:46)

ewaller · 2014-09-26 16:40:42

cmtonkinson wrote:

Off-topic: Since when has the forum search here on bbs.archlinux.org been rate limited to one search per 30 seconds?

IIRC, this applies when you are not logged in.

ninian · 2014-09-26 18:01:57

In Settings / System, is the following option ticked?

Use hardware acceleration when available

I've had to switch this option off on several Intel systems because I couldn't get the cursor changing when hovering over weblinks.

cmtonkinson · 2014-09-26 18:08:04

ninian wrote:

In Settings / System, is the following option ticked?
Use hardware acceleration when available
I've had to switch this option off on several Intel systems because I couldn't get the cursor changing when hovering over weblinks.

Why yes, yes it was ticked. I disabled it, restarted the browser, and now it is properly responding to mouse movements, which is a shame because the performance is noticeably degraded without acceleration.

So, that's one problem solved. Any idea whether this is Chromium issue or something in the video stack?

~ $ yaourt -Q | grep 'mesa\|xf86'
multilib/lib32-libxxf86vm 1.1.3-1
multilib/lib32-mesa 10.2.8-1
multilib/lib32-mesa-libgl 10.2.8-1
extra/libxxf86vm 1.1.3-1
extra/mesa 10.2.8-1
extra/mesa-libgl 10.2.8-1
extra/xf86-input-evdev 2.9.0-2 (xorg-drivers xorg)
extra/xf86-video-ati 1:7.4.0-3 (xorg-drivers xorg)
extra/xf86-video-fbdev 0.4.4-3 (xorg-drivers xorg)
extra/xf86-video-modesetting 0.9.0-2 (xorg-drivers xorg)
extra/xf86vidmodeproto 2.3.1-3

ninian · 2014-09-26 18:15:36

cmtonkinson wrote:

I disabled it, restarted the browser, and now it is properly responding to mouse movements, which is a shame because the performance is noticeably degraded without acceleration.
So, that's one problem solved. Any idea whether this is Chromium issue or something in the video stack?

Have a look at this thread too and it sounds like it could be the new GUI toolkit called Aura to blame.

cmtonkinson · 2014-09-26 18:18:41

Update:

Having identified the solution, tracking the root cause wasn't insanely difficult.

First, I was able to identify issue #313427 on the Chromium project, which traces back to an interaction between Chromium and the real culprit here, which is Synergy, which is outlined in bug #3819.

Thanks for the help!

ninian · 2014-09-26 21:07:23

Ah yes, that's heading in the right direction, except I'm not running Synergy! Also I saw Issue 386015 is more recent. Seems that the problem goes wider than just Synergy as others without Synergy are finding it too.

ninian · 2015-01-21 19:34:09

Rats, here we go again - hover problem has resurfaced in Chromium version 39.0.2171.99, using Intel graphics and the i915 module.
Again, disabling hardware acceleration makes it work properly.
Hate regressions like this ...

Arch Linux

#1 2014-09-26 16:38:54

[solved] Hover broken with Chromium

#2 2014-09-26 16:40:42

Re: [solved] Hover broken with Chromium

#3 2014-09-26 18:01:57

Re: [solved] Hover broken with Chromium

#4 2014-09-26 18:08:04

Re: [solved] Hover broken with Chromium

#5 2014-09-26 18:15:36

Re: [solved] Hover broken with Chromium

#6 2014-09-26 18:18:41

Re: [solved] Hover broken with Chromium

#7 2014-09-26 21:07:23

Re: [solved] Hover broken with Chromium

#8 2015-01-21 19:34:09

Re: [solved] Hover broken with Chromium

Board footer