long (25 seconds) pause when starting hal-device-manager
Artem.Kachitchkin at Sun.COM
Wed Mar 1 16:37:10 PST 2006
John (J5) Palmieri wrote:
> I've further pinpointed the issue and we are working on resolving it in
> dbus. http://lists.freedesktop.org/archives/dbus/2006-March/004390.html
> On Wed, 2006-03-01 at 18:07 -0500, John (J5) Palmieri wrote:
>> So it seems my patch did not fix the issue. It is really weird due to
>> the fact we get stuck in poll once consistently and then all subsequent
>> calls just go happily on their way after the poll times out. It really
>> has been a pita to debug. I know what causes it. We fire off an
>> introspect under the hood and then block for it to return if a blocking
>> call is made. This works fine for three of the calls which introspect
>> the bus but we block on the first introspect to HAL. Funny thing is if
>> we don't connect to the signals (the cause of the first three
>> introspects) everything works fine. Also if we block for introspect
>> immediately, instead of waiting for a blocking call nothing bad happens.
>> Workaround is to add result.block() to the end of the _Introspect
>> method in proxies.py.
>> On Wed, 2006-03-01 at 11:38 -0800, Artem Kachitchkine wrote:
>>> Hey John,
>>>>> I think this is a well-known problem with introspection and the python
>>>>> bindings. John?
>>>> I have a fix I need to commit.
>>> Have this been committed?
>>> I'm seeing a much longer delay on Solaris, actually never seen it time
>>> out even after two hours wait. I'd rather start debugging with your fix,
>>> since I'm not a Python expert and it'd be a significant time investment
>>> for me.
More information about the hal