Chrome/Chromium browser stuck in front (_NET_ACTIVE_WINDOW)

[Hybrid512]

Hi,

With Chrome/Chromium browser > v49, there is a very annoying bug when running on MATE Desktop in an X2Go session.

Bug is highly reproducible and occures on Ubuntu 16.04/MATE 1.12 and Ubuntu 18.04/MATE 1.20 with latest NX-libs (3.5.99.16).

How to reproduce :

• Start a windowed or fullscreen remote MATE X2Go session
• Install Chrome/Chromium web browser
• Open Chrome/Chromium and maximize it
• Open any other application that opens a window in front of the maximized Chrome/Chromium window
• Without doing anything on the new opened app, just click on Chrome/Chromium window to bring it front

From that moment, you can't cycle anymore between applications (Alt+tab) nor use the opened windows list applet.

In .xsession-error, you can see that kind of errors popping :

Window manager warning: last_user_time (6169163) is greater than comparison timestamp (2537489855). This most likely represents a buggy client sending inaccurate timestamps in messages such as _NET_ACTIVE_WINDOW. Trying to work around... Window manager warning: 0x2000001 (Google - C) appears to be one of the offending windows with a timestamp of 6169163. Working around... Window manager warning: last_user_time (6179515) is greater than comparison timestamp (2537496185). This most likely represents a buggy client sending inaccurate timestamps in messages such as _NET_ACTIVE_WINDOW. Trying to work around... Window manager warning: 0x2000001 (Google - C) appears to be one of the offending windows with a timestamp of 6179515. Working around... Window manager warning: last_user_time (6201788) is greater than comparison timestamp (2537520141). This most likely represents a buggy client sending inaccurate timestamps in messages such as _NET_ACTIVE_WINDOW. Trying to work around... Window manager warning: 0x2000001 (Google - C) appears to be one of the offending windows with a timestamp of 6201788. Working around...

This only happens with nxagent and not with Xephyr (Sunweaver's note) This doesn't occure with XFCE, only MATE remotely through X2Go but not on local.

[uli42]

Looks like this bug: https://bugs.freedesktop.org/show_bug.cgi?id=39367 There's a patch for libX11 included. Must check if we have that in our libX11 (probably yes). We should also check if it is included in the system's libX11 of the affected systems.

[uli42]

Can you please change the window manager in the Mate session to something else and retry)?

[uli42]

Chrome complains about missing GLX 1.3 in the nx session. So how to you manage getting it started?

[Hybrid512]

Chrome adapts itself when it lacks any feature/extension quite well. This is true that it complains at startup but it starts anyway. I created a policy with these values to mute these error messages :

• "HardwareAccelerationModeEnabled": false
• "Disable3DAPIs": true

But for sure, having a more up to date GLX support would probably help but I doubt this is related to this specific problem.

[Hybrid512]

Can you please change the window manager in the Mate session to something else and retry)?

How can I and which WM do you propose ?

[uli42]

I don't know MATE in detail. But generally you can replace a running WM by starting another one with --replace. Try openbox for example.

Regarding that freedesktop bug: our libNX_X11 has the patches from the that bug incorporated.

I have not been able to start chrome due to OpenGL issues when I tried last time. Can you give me a pointer how to setup a chroem profile as you mentioned above?

[uli42]

I have been unable to trigger this behaviour on Ubuntu 16.04 with

• google-chrome-stable 57.0.2987.133-1
• mate-desktop 1.12.1-1

I could run chrome with a new policy (google-chrome --user-data-dir=~/.config/google-chrome-test). Unfortunately I could not find the hardware acceleration and 3D settings you mentioned above. So I simply disabled "use hardware acceleration is possible".

My nxagent was 3.5.99-16 with this BIG-REQUESTS patch: https://github.com/ArcticaProject/nx-libs/pull/729

So can you please

[uli42]

Further research show that it is probably to some issue in gtk and/or metacity. Many gtk apps are showing these "buggy client" messages.

It could also be that these messages are totally unrelated to the "chrome stays on top" problem.

Some things to test:

• run chrome with "use system's window decorations"
• compile chromium without this patch: https://chromium.googlesource.com/chromium/src.git/+/ca3656dc321be956db788c5fc8ee16cd20ed6d86

[uli42]

@Hybrid512 can you confirm that the problem is not happening with a current chrome? (what version)?

[Hybrid512]

nope, this is still happening In rare cases, it doesn't appear but in a non reproducible manner. However, it happens nearly 99% of the time, whatever recent Chrome/Chromium release and whatever MATE release (tested from ubuntu 16.04/MATE 1.12 to Ubuntu 18.04/MATE 1.20). BTW, this is strictly a MATE bug because I don't have the issue with XFCE but it only occurs under X2Go and not on a local session on real hardware.

[taizel]

Any new on this? I confirm I suffer from the same issue, running MATE on Mint (based on Ubuntu 18.04.3). Same details, it only occurs under X2Go and not on a local session on real hardware.

[Hybrid512]

As I said, this is a MATE bug ... I switched a long time ago to XFCE, while still using many MATE applications (Pluma, EOM, ....) it works way better and is very stable and fast. I liked MATE a lot but unless there is a fix in a recent release, I would recommend switching to XFCE.

[uli42]

I have added some fixes (contained in 3.5.99.25) for the mutter window manager, which stems from MetaCity, just as MATE's Marco does. Maybe the behaviour is better now. Someone needs to check this.

On Thu, Nov 12, 2020 at 3:23 PM Walid Moghrabi [email protected] wrote:

As I said, this is a MATE bug ... I switched a long time ago to XFCE, while still using many MATE applications (Pluma, EOM, ....) it works way better and is very stable and fast. I liked MATE a lot but unless there is a fix in a recent release, I would recommend switching to XFCE.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/ArcticaProject/nx-libs/issues/697#issuecomment-726108582, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABQHBZGKQFCPKXNJWHQ3ECTSPPVUZANCNFSM4FBXUKMQ .