Guys,

Here is a new issue on the latest Arch with nvidia and any gtk app like
Firefox or Thunderbird. When using Firefox or Thunderbird, X cpu usage
skyrockets to 85+%, the Tbird cpu will be 20+%. Here is a top screenshot
showing usage:


http://www.3111skyline.com/dl/arch/bugs/tbird-cpu.jpg

Merely scrolling the new messages window or a web page causes X cpu usage
to jump from a resting usage of 1-5% to over 30%. This is viewing all messages
as plain text only. When tbird checks imap servers or sends messages both X
and tbird CPU use goes way way up. Anybody else seeing something similar?


gtk 1.2.10-11
gtk2 2.24.11-2
gtk3 3.4.4-1

linux 3.4.7-1
nvidia 302.17-4

firefox 14.0.1-1
thunderbird 14.0-1

xorg-server 1.12.3-1
xorg-server-common 1.12.3-1

It's like there is some very bad race condition created in X when using gtk
apps with the nvidia driver?


--
David C. Rankin, J.D.,P.E.

On 08/02/2012 04:00 PM, David C. Rankin wrote:
>
> It's like there is some very bad race condition created in X when using gtk
> apps with the nvidia driver?

For what it is worth, I tested a SuSE box with
MozillaThunderbird-14.0-2.1.x86_64 and xorg-x11-7.6-227.1.x86_64. The combined
CPU usage of tbird and X never exceeds 40%. (radeonhd driver)

Let me know if I can send anything else from the Arch box exhibiting the high
cpu usage.

--
David C. Rankin, J.D.,P.E.

On Thu, Aug 2, 2012 at 10:37 PM, David C. Rankin
<drankinatty@suddenlinkmail.com> wrote:
> On 08/02/2012 04:00 PM, David C. Rankin wrote:
>>
>> It's like there is some very bad race condition created in X when using gtk
>> apps with the nvidia driver?

Are you running ntpd ? Could be related to:
http://serverfault.com/questions/403732/anyone-else-experiencing-high-rates-of-linux-server-crashes-during-a-leap-second

Try stopping the ntpd and setting the date manually as suggested. Then
start ntpd again.

I had this problem on a Java application (activeMQ)

But it could be irrelevant also..
--
Caution: breathing may be hazardous to your health.

#include <stdio.h>
int main(){printf("%s","x4cx65x6fx6ex69x64x61x73");}

On 08/02/2012 05:24 PM, Leonidas Spyropoulos wrote:

Are you running ntpd ? Could be related to:
http://serverfault.com/questions/403732/anyone-else-experiencing-high-rates-of-linux-server-crashes-during-a-leap-second

Try stopping the ntpd and setting the date manually as suggested. Then
start ntpd again.

I had this problem on a Java application (activeMQ)



I was bitten by this as soon as the leap second happened and followed the
thread "Re: [arch-general] Leap seconds ntp and chrony?" I host a couple of
minecraft servers for my kids. As soon as the leap second was injected, the
apps crashed dead in their tracks. That was a completely different box though.


Here, time itself doesn't seem to be the issue. The sysclock and hwclock
are fairly well synced:


09:46 providence:~> date && sudo hwclock --show
Fri Aug 3 09:46:19 CDT 2012
Fri 03 Aug 2012 09:46:20 AM CDT -0.736355 seconds

The drift does bounce a bit, but I'm not sure that matters:

09:46 providence:~> date && sudo hwclock --show
Fri Aug 3 09:46:19 CDT 2012
Fri 03 Aug 2012 09:46:20 AM CDT -0.736355 seconds

09:46 providence:~> date && sudo hwclock --show
Fri Aug 3 09:47:21 CDT 2012
Fri 03 Aug 2012 09:47:22 AM CDT -0.891175 seconds

09:47 providence:~> date && sudo hwclock --show
Fri Aug 3 09:47:28 CDT 2012
Fri 03 Aug 2012 09:47:29 AM CDT -0.250572 seconds

I'll keep and eye on this issue. I wonder if any of the latest changes to
the init/shutdown system or rc.conf could have done this?


--
David C. Rankin, J.D.,P.E.