Extremly long login time in F15?

[Panu Matilainen]

On 03/30/2011 11:05 AM, Joachim Backes wrote:
> On 03/30/2011 10:02 AM, Panu Matilainen wrote:
>> On 03/30/2011 10:22 AM, Joachim Backes wrote:
>>> On 03/30/2011 09:14 AM, Panu Matilainen wrote:
>>>> On 03/30/2011 08:07 AM, Joachim Backes wrote:
>>>>> Hi,
>>>>>
>>>>> I'm running F15 with gnome-shell.
>>>>>
>>>>> Having the following effect: after having entered the login
>>>>> password, it
>>>>> takes about 10-15 secs until the session is running and the
>>>>> gnome-shell
>>>>> panel appears at the desktop's top. For me, that extreme long: I
>>>>> have a
>>>>> dual core Intel CPU, both with 1.8 Ghz, and 2 Gigs of mem.
>>>>>
>>>>> Somebody can explain this effect?
>>>>
>>>> I'm seeing this too, both in Gnome 3 and XFCE sessions (where
>>>> especially
>>>> the latter should fire up in a blink of an eye), so it's something else
>>>> than gnome-shell as such. Haven't got around to look into it (yet)
>>>> though. Oh and no network filesystems involved here either.
>>>>
>>>> - Panu -
>>>
>>> Hi Panu,
>>>
>>> I took a look in /var/log/messages, and there is a time leap of 30 secs
>>> between
>>>
>>> Mar 30 08:57:03 ..... gdm-simple-greeter[6636]: DEBUG(+):
>>> GdmGreeterSession: Disposing greeter_session
>>>
>>> and
>>>
>>> Mar 30 08:57:33 ..... gdm-simple-slave[6573]: DEBUG(+): GdmCommon:
>>> process (pid:6597) done (status:0)
>>>
>>> Do you see similar things?
>>
>> Yes, very much the same, there's a 30 sec pause in the log:
>>
>> [...]
>> Mar 30 10:45:48 turre gdm-simple-slave[2160]: DEBUG(+):
>> GdmWelcomeSession: Stopping welcome_session
>> Mar 30 10:45:48 turre gdm-simple-slave[2160]: DEBUG(+): GdmCommon:
>> sending signal 15 to process 2184
>> Mar 30 10:45:48 turre gdm-simple-slave[2160]: DEBUG(+):
>> GdmWelcomeSession: Waiting on process 2184
>> Mar 30 10:45:48 turre gnome-session[2184]: WARNING: Invalid method call:
>> Argument 0 is specified to be of type "boolean", but is actually of type
>> "uint32"#012
>> Mar 30 10:45:50 turre gnome-session[2184]: WARNING: Client
>> '/org/gnome/SessionManager/Client2' failed to reply before timeout
>> Mar 30 10:45:50 turre gnome-session[2184]: WARNING: Invalid method call:
>> Argument 0 is specified to be of type "boolean", but is actually of type
>> "uint32"#012
>> Mar 30 10:45:50 turre gdm-simple-greeter[2219]: DEBUG(+):
>> GdmGreeterSession: Disposing greeter_session
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+): GdmCommon:
>> process (pid:2184) done (status:0)
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+):
>> GdmWelcomeSession: WelcomeSession died
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+):
>> GdmWelcomeSession: De-registering session from ConsoleKit
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+):
>> GdmWelcomeSession: Stopping D-Bus daemon
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+): GdmCommon:
>> sending signal 15 to process -2182
>> Mar 30 10:46:20 turre gdm-simple-slave[2160]: DEBUG(+): GreeterServer:
>> Stopping greeter server...
>> [...]
>>
>> So quite apparently something is misfiring and there's a 30sec timeout
>> on something, somewhere...
>>
>> Also I /think/ this didn't happen in F15-alpha originally when I
>> installed it,
>
> Hi Panu,
>
> I can confirm this. This behaviour appeared after some updates. I don't
> know what updates are the culprit :-(

Turns out the bug is in at-spi2-core, exposed by a change in the way gdm 
runs its login-screen gnome-session. Was fun to dig up :)

For the impatient, 'yum remove "at-spi2*"' should do the trick until an 
update is built for https://bugzilla.redhat.com/show_bug.cgi?id=691995.

	- Panu -